Send us your content and instructions. Get a film, a course, or an LMS package back.

A contract for a client application that supplies its own media — HeyGen avatar clips, Higgsfield videos, images — and wants it stitched, captioned, branded, and packaged. The engine is a compositor: you provide the assets, it assembles them.

You provide the media — no generation keys needed Linear video · branching course · SCORM / xAPI / cmi5 JSON over HTTPS · async jobs

✦ What Studio can do (and how to ask)

Every capability, a sentence to say it, and the request to send

Find it here, say it in plain words, then send the request. The live engine is version 3.69.0 (GET /health reports it). Items marked new in 3.69.0 arrived in this release; items without a badge existed before. Every request needs Authorization: Bearer mk_live_…; a POST returns a jobId to follow at GET /jobs/{jobId}.

Docs and checks: /docs (this guide) · /llms-full.txt (the complete reference for app builders) · /llms.txt · GET /v1 (the live index, needs your key) · /health (version and what is on, no key). Leave a choice out and Studio picks (look, music, voice, set, frame); the finished job records what it picked.

1. Make a film

Film from a brief (Director)
Describe the film in plain words; Studio writes the script, makes the pictures, voices it, scores it and renders it (usually 3 to 6 minutes).
Say it: “Make a 30-second vertical trailer for my novel The Salt Road, warm and unhurried, end on the title and 'Out now'.”
request
POST /v1/direct
{
  "brief": "A 30-second trailer for 'The Salt Road', a novel about a family of salt traders. For Instagram readers. Warm, unhurried. End on the title and 'Out now'.",
  "title": "The Salt Road",
  "preset": "9:16",
  "target": {
    "seconds": 30
  },
  "generateVisuals": true
}
Other fields it reads: assets [{ id, kind: image|video|text|music, src, note }], theme, tone, avoid, narration (false = no voice), voiceId, voiceProvider, language, score, music, brand, headings, artStyle (ink | ink-mono), imageProvider (openai | google), motion (flow | flow-left | flow-right), style (blocks), webhook, priority. Defaults: preset 16:9; theme left out or 'auto' = Studio picks the look that fits the brief; narration on when a voice is available; music/score chosen from the tone. Poll GET /jobs/{id}: storyboard appears first (DELETE /jobs/{id} stops before the costly part); done gives url (MP4), posterUrl, captionsUrl (SRT), subtitlesUrl (VTT), chaptersUrl, film, notes[], picked and look.
Film from my own clips and pictures (timeline)
You supply the media; Studio stitches, captions, brands and mixes it. Scenes take video, image (Ken Burns), color cards, 3d scenes or blocks.
Say it: “Join these two clips and this slide into a 16:9 film with a title over the first one and a fade between them.”
request
POST /v1/render
{
  "timeline": {
    "preset": "16:9",
    "theme": "lux",
    "scenes": [
      {
        "duration": 6,
        "visual": {
          "type": "video",
          "src": "https://your-cdn/intro.mp4"
        },
        "transition": "fade",
        "overlays": [
          {
            "type": "title",
            "text": "Welcome"
          }
        ]
      },
      {
        "duration": 5,
        "visual": {
          "type": "image",
          "src": "https://your-cdn/slide.png",
          "kenBurns": "in"
        }
      }
    ]
  }
}
Scene fields: duration (not needed when audio.src narration is given), visual, overlays, effects, atmosphere, transition, caption (burned), subtitle (SRT only), audio, highlights, marks. Timeline fields: preset, output { width, height, fps, crf }, theme, brand, music, score, sfx, subtitles, subtitleStyle, frame, director, language, bookend. Src accepts https URL, data: URL, { inline, mime } or a stored asset { __asset, mime } from POST /v1/assets. Ken Burns: in, out, drift-left, drift-right, punch, punch-in, hold. Up to 500 scenes.
Upload your own media first
Send big recordings once (raw bytes, up to 512 MB) and get back a stored source you can point any route at.
Say it: “Upload my interview video so I can caption it.”
request
POST /v1/assets
(raw file bytes as the body, with its Content-Type header; not JSON)
Returns an id and a source ({ __asset, mime }). Phone to computer: POST /v1/handoff returns a link/QR; the phone uploads to the exact slot. POST /v1/download-link { path } makes a 6-hour resumable link to a finished file.
Many films from one list (merge)
One template with {{column}} placeholders and a row per film; each row is its own fast job.
Say it: “Make a square 'Welcome, Priya' film for every person on this list.”
request
POST /v1/merge
{
  "template": {
    "preset": "square",
    "scenes": [
      {
        "duration": 4,
        "visual": {
          "type": "color"
        },
        "overlays": [
          {
            "type": "title",
            "text": "Welcome, {{first}}"
          }
        ]
      }
    ]
  },
  "rows": [
    {
      "first": "Priya"
    },
    {
      "first": "Kenji"
    }
  ],
  "key": "first"
}
Follow the set with GET /v1/batches/{batchId}. Save a template: PUT /v1/templates/{name} (each PUT is a new version); GET /v1/templates, GET /v1/templates/{name}?version=N, DELETE /v1/templates/{name}.
Personal films for a client list (campaigns)
One talking-cast script with {first_name}, {company} (fallback form {company|your team}) and a client list gives a personal film per client, each with a watch page, a moving GIF for email and open/play/click tracking.
Say it: “Send each client a personal thank-you film from Maya using their first name and company.”
request
POST /v1/campaigns
{
  "name": "Thank you",
  "set": "office",
  "cast": [
    {
      "id": "maya",
      "preset": "maya"
    }
  ],
  "lines": [
    {
      "who": "maya",
      "text": "Hi {first_name}, thank you from everyone at {company|our team}."
    }
  ],
  "clients": "first_name,company\nPriya,Acme",
  "cta": {
    "label": "Book a call",
    "url": "https://example.com/book"
  },
  "format": "vertical"
}
GET /v1/campaigns, GET /v1/campaigns/{id}, GET /v1/campaigns/{id}/export.csv (links, GIFs, email snippets, who opened). Up to 1,000 clients. clients may be a CSV string or a list. Optional brand { name, color, logo }.
Film from a song (Reel)
Lays pictures across a song's whole length with cuts on the sung phrases (chorus faster), mood-matched motion and optional timed or karaoke lyrics.
Say it: “Make a vertical lyric video from this song with my photos, dreamy mood, karaoke words.”
request
POST /v1/reel
{
  "song": "<url, data-url or stored source>",
  "visuals": [
    {
      "src": "https://.../photo1.jpg"
    },
    {
      "prompt": "sunrise over a quiet harbour"
    }
  ],
  "shape": "vertical",
  "mood": "dreamy",
  "lyrics": "auto",
  "karaoke": true,
  "captionLook": "bold"
}
mood: mellow, dreamy, romantic, dark, balanced, uplifting, epic, playful, energetic or free text / auto. shape (or shapes for several at once, up to 4), theme, title { text, subtitle }, endCard { text }, captions, lyricsText (your exact words, aligned to the audio), captionLang, fit (cover | contain), artStyle, weather, grade, review. Generated stills park at status 'review' for you to approve (POST /jobs/{id}/approve, POST /jobs/{id}/scene/{i}/regenerate).
One steady camera pan across pictures (flow camera)
Consecutive picture scenes become one continuous pan at one speed, no stops or jumps; repeat the same src to split a wide panorama.
Say it: “Pan slowly left to right across these four street photos as one continuous shot.”
request
POST /v1/render
{
  "timeline": {
    "preset": "16:9",
    "scenes": [
      {
        "duration": 4,
        "visual": {
          "type": "image",
          "src": "https://.../street.png",
          "flow": "right"
        }
      },
      {
        "duration": 4,
        "visual": {
          "type": "image",
          "src": "https://.../street2.png",
          "flow": "right"
        }
      }
    ]
  }
}
Since 3.68.0. flow is 'right' or 'left'. Image scenes only, not with highlights or marks (refused). On /v1/direct use motion: 'flow' (or 'flow-left' / 'flow-right').
Words laid out as the picture (block layout and brand kit)
The reel look where words ARE the picture: text with a small lead-in and a big phrase, slanted picture panels, one tile per spoken word, count-up numbers, each entering as it is spoken, in your brand colours.
Say it: “Make it a block-layout reel: 'If you're looking to / GROW YOUR AUDIENCE', then 'over the last / 14 years' counting up, in our maroon and gold.”
request
POST /v1/render
{
  "timeline": {
    "preset": "9:16",
    "brand": {
      "color": "#5A0F14",
      "accent": "#E2B04A",
      "endCard": {
        "text": "Follow for more"
      }
    },
    "scenes": [
      {
        "duration": 5,
        "visual": {
          "type": "blocks",
          "items": [
            {
              "kind": "text",
              "small": "If you're looking to",
              "big": "Grow your audience"
            },
            {
              "kind": "count",
              "small": "Over the last",
              "to": 14,
              "suffix": "years"
            }
          ]
        },
        "subtitle": "If you're looking to grow your audience... over the last fourteen years..."
      }
    ]
  }
}
Since 3.68.0. Item kinds: text (small, big), panel (src, label, from, cut, height), tiles (tiles: [{ label, src }]), count (to, small, suffix). Pin timing with word or at. brand: color, accent, logo, corner, font (Fontsource name), endCard { text }. On /v1/direct use style: 'blocks'.
Software walkthrough from screenshots
One image scene per step; a glowing ring lights the control when the narration names it.
Say it: “Turn these four screenshots into a how-to, ringing the Continue button on step 2.”
request
POST /v1/render
{
  "timeline": {
    "preset": "16:9",
    "scenes": [
      {
        "duration": 7,
        "visual": {
          "type": "image",
          "src": "https://.../step-2.png",
          "kenBurns": "hold"
        },
        "overlays": [
          {
            "type": "eyebrow",
            "text": "Step 2"
          }
        ],
        "highlights": [
          {
            "x": 0.73,
            "y": 0.86,
            "w": 0.065,
            "h": 0.056,
            "start": 0.5,
            "label": "Continue"
          }
        ]
      }
    ]
  }
}
x, y, w, h are fractions of the screenshot; start/end are fractions of the scene; up to six per scene; color overrides the ring. Send screenshots at 2x the film size.
Pictures laid over a talk (enrich)
Listens to a talk, finds where a picture helps and lays it over the speaker as a full cutaway or a corner inset; the speaker's audio is untouched. Always stops for your review first.
Say it: “Add pictures over this talk where she names something concrete.”
request
POST /v1/enrich
{
  "video": "<stored source>",
  "source": "ai",
  "count": 6
}
source: mine (your pictures[] in order) | ai | both. kind, captions, captionLook, captionLang, spokenLang, review: false to skip the review stop. Approve with POST /jobs/{id}/approve.
Plan only, no render (storyboard)
Gives back the beats of a script or a recording without rendering, and draws one picture per beat.
Say it: “Show me the storyboard for this text before we make anything.”
request
POST /v1/storyboard
{
  "text": "Our school opens a new library on Monday..."
}
Send text, or asset (a stored recording) with optional language. One picture: POST /v1/storyboard/image { prompt (or visualPrompt), preset, artStyle, avoid, keep: true }.
Recorded clips and cast scenes in one film (mix), with pause-and-explain and shows
Blocks in order: recorded clips (trimmed with from/to, keeping their sound) and cast scenes, one set of captions. A clip's stops freeze the recording while characters explain, with a ring drawn round what is talked about. A show block puts a host and guests in windows.
Say it: “Play my screen recording, pause at 4 seconds and have Grace explain the button, circling it.”
request
POST /v1/mix
{
  "format": "landscape",
  "blocks": [
    {
      "type": "clip",
      "src": "<stored source>",
      "stops": [
        {
          "at": 4,
          "side": "right",
          "cast": [
            {
              "id": "grace",
              "preset": "grace"
            }
          ],
          "lines": [
            {
              "who": "grace",
              "text": "This is the button to press.",
              "mood": "happy"
            }
          ],
          "circles": [
            {
              "find": "the blue button",
              "cue": "button"
            }
          ]
        }
      ]
    }
  ]
}
Show blocks need 'show' in GET /health features: { type: 'show', sources: { mei: { kind: 'cast', preset: 'mei' }, expert: { kind: 'clip', src, captions: 'auto' } }, beats: [{ who, text, layout: host | host-with-guest | guest-big | guest-full | nest | side-by-side, guest, zoom }] }. The job's result.qc[] is Studio's own check of the show.
A generated video clip (Veo) - owner only
Asks for a price first; the same body with confirm: true makes an 8-second clip. Off unless the engine operator turns it on; capped.
Say it: “(Operator only) How much for a clip of a woman saying hello in a kitchen?”
request
POST /v1/videogen
{
  "parts": [
    "A woman in a kitchen says: Hello, welcome."
  ],
  "aspectRatio": "16:9"
}
Add confirm: true to make it. GET /v1/videogen shows the budget.
Best moments from a long video (shorts) - invite only
Finds the best 20-90 second moments with a score, hook, title and cost estimate; nothing is cut. Only the engine owner can start it for now; any other key gets 403 invite_only.
Say it: “Find the 8 best moments in this hour-long interview for shorts.”
request
POST /v1/moments
{
  "video": "<stored source>",
  "count": 8,
  "minSeconds": 20,
  "maxSeconds": 60
}
Also from (a jobId, to find more without re-listening), spokenLang, names, words: false. Read GET /v1/moments/{jobId} and GET /v1/moments/{jobId}/transcript?format=json|txt|srt. Up to 240 minutes.
Control a running job
Stop, re-run, report and follow jobs.
Say it: “Cancel that film. / Run it again. / The check is wrong, tell the team.”
request
DELETE /jobs/{jobId}
POST /jobs/{jobId}/restart re-runs from retained inputs; GET /jobs/{jobId}/events streams progress (SSE); GET /v1/jobs?endUser=<customer> lists one customer's films; POST /v1/jobs/{jobId}/report { message, kind: check-wrong | film-problem | other, advisory, scene, contact } opens a ticket (GET /v1/tickets, GET /v1/tickets/{id}, POST /v1/tickets/{id}/reply).
Standing rules for every film
Your non-negotiables ('no music', 'never show faces') stored once and merged into every generation on your key.
Say it: “Always keep the pictures free of people.”
request
PUT /v1/rules
{
  "rules": "Never show faces. No music."
}
GET /v1/rules reads them back; an empty string clears them.

2. Captions and look

Burn captions into any video
Listens, writes the words, times them to the voice and burns them in; your sound is copied untouched; an SRT and a VTT come with it.
Say it: “Caption this video in bold white text.”
request
POST /v1/caption
{
  "video": "<stored source or URL>",
  "captionLook": "bold",
  "spokenLang": "en"
}
captionLook: bold (default for reels), clean, neon, cinematic, subtitle, band (band is new in 3.69.0). captionLang translates (see Voices and languages). karaoke: true lights each word as it is sung. lyricsText = your exact words aligned to the audio. Room-tone 'Thank you for watching' inventions are dropped and reported.
Band captions with one gold key word new in 3.69.0
Two or three spoken words at a time in bold capitals at the top safe area, one word per beat in gold. The gold word is a word you wrote in ALL CAPS, else a number or rank, else a proper noun, else the longest meaningful word.
Say it: “Use band captions with a gold key word.”
request
POST /v1/render
{
  "timeline": {
    "preset": "9:16",
    "subtitleStyle": {
      "look": "band",
      "wordsPerBeat": "auto",
      "keyWord": "auto",
      "accent": "#D9A441",
      "caseMode": "upper"
    },
    "subtitles": [
      {
        "start": 0,
        "end": 3,
        "text": "We grew 40% in one year",
        "words": [
          {
            "word": "We",
            "start": 0,
            "end": 0.2
          }
        ]
      }
    ],
    "scenes": [
      {
        "duration": 6,
        "visual": {
          "type": "video",
          "src": "https://.../me.mp4"
        }
      }
    ]
  }
}
New in 3.69.0. Also captionLook: 'band' on /v1/caption, /v1/reel, /v1/enrich. wordsPerBeat: auto or 1-4; keyWord: auto | caps | off; caseMode: upper | as-written. Beats follow speech when cues carry words [{ word, start, end }]; otherwise spread evenly and the result says so. Result: picked.captions.
Pop captions (karaoke word-pop)
One word at a time, big on a plate, springing in; write the key word in ALL CAPS and it takes the accent colour.
Say it: “Pop the words on screen one at a time, with WIN in the accent colour.”
request
POST /v1/render
{
  "timeline": {
    "scenes": [
      {
        "duration": 4,
        "visual": {
          "type": "color"
        },
        "overlays": [
          {
            "type": "popcaption",
            "text": "you can WIN this game",
            "start": 0.5,
            "end": 3.5
          }
        ]
      }
    ]
  }
}
wordsPerBeat groups words; emphasis: false opts out of the caps rule.
Your own captions, bilingual and complex scripts
Timeline-level cues burned through libass, so Tamil, Arabic, Devanagari and other complex scripts shape correctly; add a second language line.
Say it: “Burn these captions in, English with Tamil underneath.”
request
POST /v1/render
{
  "timeline": {
    "language": "en",
    "subtitles": [
      {
        "start": 0.3,
        "end": 3.8,
        "text": "We love the land we live in",
        "ta": "நாங்கள் வாழும் நிலத்தை நேசிக்கிறோம்"
      }
    ],
    "scenes": [
      {
        "duration": 5,
        "visual": {
          "type": "color"
        }
      }
    ]
  }
}
subtitleStyle overrides: enSize, taSize, enColor, taColor, latinFont, tamilFont, marginV, outline, shadow. captions: { subtitles: [...] } is the same thing. An SRT and VTT sidecar are always written.
Look: themes, art styles, formats
The film's visual voice and shape. Themes: lux, noir, pulse (dark); ivory, azure, sunlit, bloom, meadow, candy, ember (bright). Art styles: ink, ink-mono. Formats: 16:9, 9:16, 1:1.
Say it: “Make it bright and sunny, ink-comic pictures, vertical.”
request
POST /v1/direct
{
  "brief": "A bright 20-second explainer on how rainbows form",
  "theme": "sunlit",
  "artStyle": "ink",
  "preset": "9:16"
}
Leave theme out or 'auto' and Studio picks for the topic (never dark by default). Presets also accept aliases: landscape, course, lecture, webinar | portrait, vertical, social, tiktok, reels, shorts, story | square, feed, post. Exact size with output { width, height, fps, crf } (even numbers, up to 4K). Lists: GET /v1/themes (themes, artStyles, lessonLooks), GET /v1/presets.
Grades, textures and transitions
Per-scene cinematic looks, weather and light layers, and scene-to-scene transitions.
Say it: “Give scene 2 a golden-hour grade with film grain and falling snow, and whip into the next scene.”
request
POST /v1/render
{
  "timeline": {
    "scenes": [
      {
        "duration": 5,
        "visual": {
          "type": "image",
          "src": "https://.../a.jpg",
          "kenBurns": "drift-left"
        },
        "effects": [
          "golden",
          {
            "name": "grain",
            "intensity": 0.4
          }
        ],
        "atmosphere": [
          {
            "effect": "snow",
            "density": 0.6
          }
        ]
      },
      {
        "duration": 4,
        "visual": {
          "type": "color"
        },
        "transition": {
          "type": "whip",
          "duration": 0.6
        }
      }
    ]
  }
}
Effects: golden, warm, cool, moonlight, cinematic, noir, sepia, vibrant, faded, grain, vignette, chroma, sharpen, soft, dream, backdrop, vintage, flash. Atmosphere: snow, embers, fog, light (or your own overlay clip { src, blend: 'screen' }). Transitions: cut, fade, fadeblack, dissolve, slides, wipes, circleopen, shatter, whip, slice, mosaic, punch, and soft ones breathe, veil, curtain, bloom, fadeslow. Lists: GET /v1/effects, GET /v1/atmosphere. Match the turn to the tone; /v1/direct takes tone to steer it.
Text on screen (overlays) and scene headings
Eyebrow, title, subtitle, lower-third, keyline, caption, popcaption and lockup overlays; kinetic and spring entrances; scene headings that place themselves clear of the subject.
Say it: “Put 'Jane Doe, Founder' as a lower third at 0.5 seconds, and head each scene automatically.”
request
POST /v1/render
{
  "timeline": {
    "scenes": [
      {
        "duration": 5,
        "visual": {
          "type": "video",
          "src": "https://.../clip.mp4"
        },
        "heading": "The referral engine",
        "overlays": [
          {
            "type": "lowerthird",
            "text": "Jane Doe",
            "secondary": "Founder, ACME",
            "start": 0.5,
            "end": 5
          },
          {
            "type": "lockup",
            "eyebrow": "Lesson",
            "title": "Safety basics",
            "kinetic": true,
            "animate": "spring"
          }
        ]
      }
    ]
  }
}
On /v1/direct send headings: true so the Director labels each scene. Text is sized and wrapped for the frame, and kept clear of platform buttons on vertical films.
Frame balance: fill empty zones with designed blocks new in 3.69.0
Looks at each picture or video scene, finds the subject, measures the empty zones and puts one small designed block in an empty one (a block caption, a counting-up number, a label pill or a tile); never over the subject.
Say it: “Balance the frame and fill any empty space with block captions and a counter.”
request
POST /v1/render
{
  "timeline": {
    "frame": {
      "balance": "auto",
      "fill": [
        "blockcaption",
        "counter"
      ],
      "sound": "soft"
    },
    "scenes": [
      {
        "duration": 6,
        "visual": {
          "type": "video",
          "src": "https://.../me.mp4"
        }
      }
    ]
  }
}
New in 3.69.0. Off unless frame.balance is 'auto'. Per scene: scene.frame.balance: 'off'. Hand-placed: overlays { type: 'counter', value, suffix, label, box }, { type: 'blockcaption', small, big, accent }, { type: 'pill' | 'tile', text }. At most one per seven seconds, none in the first half second. Result: picked.frame.
Designed layout and smart reframe new in 3.69.0
Dresses each picture or video scene as a layout: a brand-colour band at the top, the picture below, and a lower zone with big two-tier phrase blocks on diagonal plates and counters. Smart reframe crops a landscape picture so the person fills the frame (eyes about 38% down) along a smooth path; screenshots and diagrams stay whole as a card.
Say it: “Use the designed layout in our maroon and gold, and make the speaker fill the frame.”
request
POST /v1/render
{
  "timeline": {
    "frame": {
      "layout": "designed"
    },
    "brand": {
      "color": "#5A0F14",
      "accent": "#E2B04A"
    },
    "scenes": [
      {
        "duration": 7,
        "visual": {
          "type": "video",
          "src": "https://.../me.mp4"
        }
      },
      {
        "duration": 5,
        "visual": {
          "type": "image",
          "src": "https://.../proof.png"
        },
        "frame": {
          "focus": {
            "x": 0.05,
            "y": 0.3,
            "w": 0.9,
            "h": 0.3
          }
        }
      }
    ]
  }
}
New in 3.69.0. frame.layout: plain | designed; frame.reframe: auto | off; frame.blocks: auto | plates | stack; frame.faceBlocks: 'sparse' for the older look; scene.frame.fit: auto | fill | card; scene.frame.focus { x, y, w, h }; scene.frame.tile { src or scene, label, start, end }; scene.frame.phrases [{ big, small, accent, start, end }] = 1-6 blocks you write yourself (scene seconds), which replace the ones picked from the words (the small line shrinks to fit; a long phrase wraps on whole words). Without phrases it needs cue words (words[] times) for the phrase blocks. Words in Indic scripts get no block. Result: picked.frame.designed.

3. Marks and effects (hand-drawn marks, confetti, bubbles)

Pen marks that draw themselves new in 3.69.0
A hand-drawn circle, box, arrow, underline or highlighter forms on a picture or video: a circle round a book cover, a box round '#1 New Release'. Position is fractions of the source picture, so it survives zoom and reframe.
Say it: “Circle the book cover, then draw a box round '#2 BEST SELLER' with that label.”
request
POST /v1/render
{
  "timeline": {
    "scenes": [
      {
        "duration": 5.5,
        "visual": {
          "type": "image",
          "src": "https://.../amazon.png"
        },
        "marks": [
          {
            "kind": "circle",
            "x": 0.08,
            "y": 0.2,
            "w": 0.19,
            "h": 0.46,
            "at": 0.6
          },
          {
            "kind": "box",
            "x": 0.33,
            "y": 0.53,
            "w": 0.57,
            "h": 0.06,
            "at": 2,
            "label": "#2 BEST SELLER"
          }
        ]
      }
    ]
  }
}
New in 3.69.0. kind: circle, box, arrow, underline (double: true), highlighter. Options: at, draw, hold, color, width, label, from, loops, sfx. Instead of x/y/w/h, find: 'the book cover' works only where the picture-reading service is set up. Up to 6 per scene, 12 per film; image or video scenes only; not with visual.flow. frame.marks: 'auto' turns highlights and frame.focus into marks. Plays a quiet draw sound unless sfx: false. Result: picked.marks.
Confetti, hearts and sparkles new in 3.69.0
Celebration bursts drawn in code: confetti from the top, hearts rising beside the face, sparkles round a cover.
Say it: “Throw confetti when she says 'number two'.”
request
POST /v1/render
{
  "timeline": {
    "scenes": [
      {
        "duration": 6,
        "visual": {
          "type": "video",
          "src": "https://.../me.mp4"
        },
        "bursts": [
          {
            "kind": "confetti",
            "at": 0.6
          }
        ]
      }
    ]
  }
}
New in 3.69.0. kind: confetti | hearts | sparkles; at, duration (0.8-4), x, y, w, h, intensity (0.3-2), seed, sfx. One per scene, 6 s apart, 6 per film. frame.fx: 'auto' adds a burst only where the spoken line earns one (a win gets confetti, a thank-you gets hearts).
Speech bubbles, quote card, sticker and icon pills new in 3.69.0
Bubbles that pop in one after another and never cover the face; a lower-third quote card with an underline that draws in; a tilted sticker; rounded labels with a small drawn icon (party, heart, bulb, chart, trophy, speech, sparkle, check, link, star, fire, thumbs; emoji map onto them).
Say it: “Add a 'WE DID IT!' sticker with a party icon and a '#1 TOP NEW RELEASE' pill with a trophy.”
request
POST /v1/render
{
  "timeline": {
    "scenes": [
      {
        "duration": 6,
        "visual": {
          "type": "video",
          "src": "https://.../me.mp4"
        },
        "overlays": [
          {
            "type": "sticker",
            "text": "WE DID IT!",
            "icon": "party",
            "start": 0.9,
            "end": 4.6
          },
          {
            "type": "pill",
            "text": "#1 TOP NEW RELEASE",
            "icon": "trophy",
            "start": 2.4,
            "end": 5.6
          },
          {
            "type": "bubble",
            "speaker": "Boss",
            "text": "Do this.",
            "side": "auto",
            "start": 1,
            "end": 5
          },
          {
            "type": "quotecard",
            "kicker": "MY CHAPTER",
            "text": "The Boss Said. The Team Heard.",
            "start": 1.2,
            "end": 4.8
          }
        ]
      }
    ]
  }
}
New in 3.69.0. Limits: bubble text 110 characters, quote card 90, sticker 24, pill 40; per scene 4 bubbles, 1 quote card, 2 stickers or pills. Overlay sound is off unless you give sfx (e.g. 'pop'). Characters the typeface cannot draw are left out and listed. Result: picked.fx.
Director auto: marks and effects chosen from what is said new in 3.69.0
Reads what is SAID over each picture and adds only what the words earn: a win gets confetti and a sticker made of the speaker's words; a thank-you gets hearts; a quote gets a bubble or quote card; evidence ('number 2', '40%') over a screenshot gets a pen mark and label; 'link in the comments' gets a link pill. Calm tones get no confetti.
Say it: “Dress this up automatically from what's said, but keep it tasteful.”
request
POST /v1/render
{
  "timeline": {
    "director": "auto",
    "tone": "warm",
    "scenes": [
      {
        "duration": 6,
        "visual": {
          "type": "video",
          "src": "https://.../me.mp4"
        }
      }
    ]
  }
}
New in 3.69.0. Off unless director is 'auto'. Hard caps: one celebration per 8 s (4 per film), 6 marks, 3 bubbles, 2 quote cards, 2 stickers, 3 pills, nothing in the first 1.5 s. Code rules decide first; TypeSafe (when the engine has a key) sorts unsure phrases; code decides what is drawn. Scenes you dressed yourself, flowing pictures and Indic-script words are left alone. Result: picked.director records every phrase and decision.

4. Sound and music

Music: pick, auto, none, or your own
Music is optional. /v1/render adds none unless asked; /v1/direct chooses for you when you say nothing. Studio's own written score has no licence or credit.
Say it: “Calm music under the whole film, only at the start and end.”
request
POST /v1/direct
{
  "brief": "A calm 90-second film about our garden project",
  "music": {
    "mood": "calm",
    "shape": "bookends"
  }
}
music: false | 'auto' | { mood: warm|calm|bright|serious|tender } | { track: '<id>' } | { src: '<your file>' }; add shape: 'bed' (default) or 'bookends' (introSeconds, outroSeconds), fadeIn, fadeOut, underVoiceLu, allowCredit. score: warm | calm | bright | serious | tender | false. Your music always wins over score. Lists: GET /v1/music (the licensed pack ships empty today), GET /v1/score. The job's musicPick says what was chosen and why; film.music is the receipt.
Small sound effects new in 3.69.0
In-house pings, pops, a draw sweep, swooshes, ticks, sparkles and a thanks chime; no licence needed.
Say it: “A soft ping when each caption pops up.”
request
POST /v1/render
{
  "timeline": {
    "sfx": "auto",
    "scenes": [
      {
        "duration": 5,
        "visual": {
          "type": "color"
        },
        "overlays": [
          {
            "type": "popcaption",
            "text": "Grow your AUDIENCE",
            "start": 0.5,
            "sfx": "ping"
          }
        ],
        "sfx": [
          {
            "name": "draw",
            "at": 2
          }
        ]
      }
    ]
  }
}
New in 3.69.0. GET /v1/sfx lists ping, ping-low, pop, draw, swoosh, tick, sparkle, thanks-chime. Ask on an overlay (sfx: 'ping' or { name, at, volume }), on a scene (sfx: [{ name, at }]) or on the timeline (sfx: [{ name, at }] or 'auto'). volume is dB under full level, -48 to -6, default -24. Nothing plays unless asked. Unknown name is refused with the valid list.
Take music out, keep the voice
Strips music, instruments and noise from audio or video and returns the voice alone (a video keeps its picture).
Say it: “Remove the background music from this recording.”
request
POST /v1/isolate
{
  "media": "<stored source or URL>"
}
Result: result.vocalsUrl and result.videoUrl. Needs the engine's isolation service (GET /v1 canRightNow.removeMusic).

5. Voices and languages

Text to a spoken clip
Type a script, get it spoken; bracketed directions like [softly] are not spoken but shape the delivery. Add language to speak in another language (translated first).
Say it: “Read this welcome in a warm voice, and keep it in my history.”
request
POST /v1/voice
{
  "text": "Welcome to the course.",
  "language": "en",
  "keep": true
}
Also provider, voiceId. Returns { audio: data-url, chars }; with keep: true it returns kept { id, url }. GET /v1/voices lists the services and voices and who is speaking today. Pin a service with provider / voiceProvider; a pinned service is the only one tried.
Your history of voices and pictures
Clips and pictures you chose to keep, newest first (100 of each kind).
Say it: “Show me the voice clips I saved.”
request
GET /v1/history?kind=voice
kind: voice | picture. GET /v1/history/{id}.mp3 (?download=1), DELETE /v1/history/{id}. Pictures are kept with keep: true on /v1/storyboard/image.
Captions and speech in 40+ languages
Translate captions into 44 languages, each in its own script with the right font fetched on demand (right-to-left included).
Say it: “Caption this film in Hindi.”
request
POST /v1/caption
{
  "video": "<stored source>",
  "captionLang": "hi",
  "spokenLang": "en"
}
Codes: en es fr de it pt nl id ms sw af so ha zu ig pl tr yo vi ru uk el ta hi mr bn te kn ml gu pa si th km ar ur fa he am ka hy zh ja ko. captionLang omitted or 'auto' keeps the spoken language (karaoke then stays on). Also on /v1/reel and /v1/enrich. language on /v1/direct, /v1/voice, /v1/cast and /v1/mix sets the spoken language. New in 3.69.0, captions burned into a film in Hindi, Marathi, Bengali and the other Indic scripts are shaped correctly: the burn now uses ffmpeg's ass filter, which joins letters and puts vowel signs on the right side. Say spokenLang: 'ur' for Urdu so it is written in Urdu script, not Hindi. Live list: GET /health captionLangs.
Indic scripts typeset correctly on covers new in 3.69.0
Headline, accent and byline in Devanagari, Bengali, Gurmukhi, Gujarati, Tamil, Telugu, Kannada and Malayalam are typeset with the script's own Noto face so vowel signs and conjuncts land correctly.
Say it: “Make the cover headline in Hindi.”
request
POST /v1/cover
{
  "from": "<finished film id>",
  "headline": "आपकी कहानी, आपकी आवाज़"
}
New in 3.69.0. picks.shaped says which lines. The badge pill cannot carry these scripts yet. Characters a font lacks (emoji) are refused in plain words.
Your customers' own avatar and copied voice
Draws a cartoon avatar from a photo and copies a voice from 10-30 seconds of speech, both with recorded consent; use them in any cast.
Say it: “Make a cartoon of me from this photo and use my voice.”
request
POST /v1/people
{
  "name": "Priya",
  "photo": "<stored source>",
  "consent": true,
  "self": true
}
Then POST /v1/people/{id}/voice { sample, consent: true }; GET /v1/people. Use in a cast as { id: 'me', avatar: 'av_...' }. With X-On-Behalf-Of each customer has their own. Someone else's avatar needs their approval (POST /v1/avatars/{id}/request); POST /v1/people/invite { name, email } makes a consent link and QR; they can withdraw at any time.

6. Characters, drawing and 3D

Talking cast (animated characters)
Characters act out your lines, voiced, lip-synced, captioned, drawn in code; under a minute to make. Up to 6 speak per scene.
Say it: “Kenji asks Grace what the book is about in the library, vertical.”
request
POST /v1/cast
{
  "set": "library",
  "format": "vertical",
  "cast": [
    {
      "id": "grace",
      "preset": "grace"
    },
    {
      "id": "kenji",
      "preset": "kenji"
    }
  ],
  "lines": [
    {
      "who": "kenji",
      "text": "What's the book about?",
      "mood": "curious",
      "action": "raise"
    },
    {
      "who": "grace",
      "text": "A family of salt traders, and one long road home.",
      "mood": "happy",
      "action": "open"
    }
  ]
}
Film at result.videoUrl. format: landscape | hd | square | vertical. style: house (default) | ink | stick. Presets: maya, zain, amira, harjit, leo, grace, sam, kenji, lucia, rosa, mei, priya, anaya, mateo. Sets include studio, classroom, park, office, library, lab, kitchen, cafe, clinic, street, beach, forest, space, stage, explainer, newsroom (full list: GET /v1/cast), or any place in words ('a volcano') or 'mine:<name>'. Moods: neutral, happy, curious, surprised, thinking, serious, sad, excited, laugh, smirk, confused, proud. Actions: explain, palm, open, raise, point, both, shrug, wave, think, count, self, big, small, ask, down. Also hold (a prop), camera, seed, gap, drawings. Hand gestures follow the words.
A topic film with a presenter
Give a topic; Studio writes a short first-person script and the avatar (or a house presenter) presents it with slides on a board.
Say it: “A 60-second film on preparing for a job interview, presented by me.”
request
POST /v1/cast
{
  "topic": "Preparing for a job interview",
  "avatar": "av_...",
  "seconds": 60
}
Optional audience, notes, set (explainer | office | newsroom | lab | cafe; left out = the one that fits the topic, shown as look). seconds 20-180. A refused topic fails with errorCode topic_declined.
Crowds of stick people
Up to 2,000 silent people behind up to 6 who speak, in the stick look: clapping, cheering, marching, queueing and more.
Say it: “A thousand people clapping in a plaza while Ana and Ben talk.”
request
POST /v1/cast
{
  "style": "stick",
  "set": "park",
  "cast": [
    {
      "id": "ana",
      "pronoun": "she",
      "age": 34
    },
    {
      "id": "ben",
      "pronoun": "he",
      "look": "ben-7"
    }
  ],
  "lines": [
    {
      "who": "ana",
      "text": "A thousand people came out today.",
      "mood": "excited"
    }
  ],
  "crowd": {
    "count": 1000,
    "seed": "plaza-a",
    "doing": "clap",
    "layout": "plaza"
  }
}
doing: stand clap wave cheer jump dance walk march sit queue (or a mix). layout: plaza stadium rows street circle scatter. Also mix of child/teen/adult/elder, palette, density, detail. Any single stick person: pronoun, age, skin, size, hair { style, color }, headwear, glasses, beard, top, bottom, look, voiceKind.
Characters and backgrounds from the shared creative library
Ask for a role or a place in words; it is reused if it exists, made if not, and kept for everyone only when generic (a nurse, a lab; never a real person, brand or film character).
Say it: “I need a tired night-shift nurse in a red top.”
request
POST /v1/library
{
  "request": "a tired night-shift nurse",
  "kind": "character"
}
kind: background | picture | character. Use it in a cast as { id: 'nurse', library: 'lib_...', top: { color: '#E53935' } }. Browse: GET /v1/library?kind=character&q=nurse. source in the answer: library | created | made-for-this-film. 20 new characters an hour per key.
Draw backgrounds, props and characters in code (free)
Describe a drawing as JSON shapes, or have Studio draw it from words or from a picture and quality-check it (up to 3 rounds), then save it and use it in any film.
Say it: “Draw me a grumpy teapot in a bowler hat I can use as a character.”
request
POST /v1/draw/create
{
  "kind": "character",
  "brief": "a grumpy teapot with a bowler hat",
  "rounds": 2,
  "save": {
    "name": "teapot"
  }
}
Also POST /v1/draw { kind, drawing, style: 'ink' } (returns a PNG), POST /v1/draw/trace { kind, picture, hint }, POST /v1/draw/check { kind, drawing, review }, POST /v1/creations { kind: background|prop|asset|character|part, name, drawing }; GET /v1/creations; PATCH and DELETE /v1/creations/{id}. Use as set: 'mine:<name>', hold: 'mine:<name>', or inline drawings { sets, props }. Invented characters: cast entry { id, drawn: {...} }; wearables: wear: [{ on: head|face|neck|chest|hips|back|hands|feet, shapes }]. GET /v1/draw lists shapes and limits.
3D scenes with real geometry
A built-in software 3D renderer: extruded 3D text, metals, glass, shadows, reflective floor, depth of field, particles, camera orbit or keyframe path, all as one scene that overlays and grades still sit on.
Say it: “A slow orbit round gold 3D letters 'ACME' on a dark reflective floor.”
request
POST /v1/render
{
  "timeline": {
    "scenes": [
      {
        "duration": 5,
        "visual": {
          "type": "3d",
          "scene": {
            "background": {
              "gradient": [
                "0x100E08",
                "0x000000"
              ]
            },
            "camera": {
              "orbit": {
                "radius": 5,
                "height": 1.3,
                "revolutions": 0.08
              },
              "fov": 38
            },
            "ground": {
              "y": -1.3,
              "reflectivity": 0.5,
              "shadow": true
            },
            "objects": [
              {
                "mesh": "text3d",
                "text": "ACME",
                "extrude": 0.4,
                "bevel": 0.03,
                "material": {
                  "color": "0xE3C878",
                  "metallic": 0.9,
                  "roughness": 0.2
                }
              }
            ]
          }
        }
      }
    ]
  }
}
Meshes: GET /v1/meshes (cube, sphere, plane, cylinder, cone, torus, icosahedron, octahedron, tetra, text, text3d, card). A single frame as a JPEG with no job: POST /v1/preview3d { scene, theme, width, height, t }. Up to 120 objects; about ten-odd seconds of render per second of 1080p. The Director stays 2D.
3D talking characters
UNCONFIRMED as a request. The 3D character builder (cast3d.mjs, character3d.mjs) exists in the engine but no route or request field exposes it; only 3D scenes (above) are reachable.
Say it: “(Do not promise this yet.)”
Treat as not available through the API until a route is confirmed.

7. Courses and learning systems (LMS, LTI)

A whole course from a brief
Studio designs the branching graph (teaching beats, scored questions, remediation loops), scripts and voices every part and renders it.
Say it: “Make a 10-minute course on kitchen fire safety with quizzes, in a clinic-style page.”
request
POST /v1/direct-course
{
  "brief": "A short course on kitchen fire safety for new cooks",
  "look": "clinic",
  "target": {
    "seconds": 600
  },
  "certificate": {
    "passingScore": 0.8
  }
}
Also nodes, tone, theme, language, narration, brandColor, celebrate, graded, previous (an earlier interactiveId: unchanged parts keep their finished films and are not re-billed), cover, outcome, author, sounds. New in 3.69.0: cover: false leaves out the covers (see Course covers), sounds: true turns on quiz sounds (see Quiz sounds and confetti). Reply { interactiveId, playerUrl }.
A branching lesson you design
Nodes are short clips ending in a question, a next node or the end; wrong answers can loop to a remediation node.
Say it: “Intro video, then a drag-and-drop question; if wrong, show a recap and ask again.”
request
POST /v1/interactive
{
  "title": "Your first bakery shift",
  "start": "intro",
  "preset": "16:9",
  "nodes": {
    "intro": {
      "timeline": {
        "scenes": [
          {
            "duration": 6,
            "visual": {
              "type": "color"
            },
            "overlays": [
              {
                "type": "lockup",
                "title": "Your first bakery shift"
              }
            ]
          }
        ]
      },
      "next": "q1"
    },
    "q1": {
      "timeline": {
        "scenes": [
          {
            "duration": 4,
            "visual": {
              "type": "color"
            }
          }
        ]
      },
      "question": {
        "type": "dragdrop",
        "prompt": "Drag each item to where it belongs.",
        "buckets": [
          "Oven",
          "Fridge"
        ],
        "items": [
          {
            "text": "Milk",
            "bucket": "Fridge"
          },
          {
            "text": "Dough",
            "bucket": "Oven"
          }
        ],
        "pass": "done",
        "fail": "intro"
      }
    },
    "done": {
      "timeline": {
        "scenes": [
          {
            "duration": 4,
            "visual": {
              "type": "color"
            },
            "overlays": [
              {
                "type": "lockup",
                "title": "Well done"
              }
            ]
          }
        ]
      }
    }
  }
}
Question types: single (default; choices [{ label, next, correct }]), multi, text, dragdrop, match, ordering, fillblanks, markwords, hotspot, numeric, and unmarked likert, poll, reflect, flashcards. Marked types route with pass and fail. explain (up to 600 characters) shows after answering. graded: true = server-graded exam mode. certificate { passingScore }. Up to 40 nodes. Poll GET /v1/interactive/{id} until status is ready (about a minute).
Lesson page look
The page around the films: journal (science, nature), clinic (health), book (children), studio, or default; free to change any time, never re-renders films.
Say it: “Switch this lesson to the picture-book look in teal.”
request
PATCH /v1/interactive/{interactiveId}
{
  "look": "book",
  "brandColor": "#0E7C7B",
  "celebrate": true
}
look 'auto' or left out picks for the topic (light unless you chose a dark theme). Preview with /play/{id}?look=book. Reply carries lessonLook { name, by, reason }.
Course covers (one per course, one per lesson) new in 3.69.0
Every ready course gets a cover for the course and one per lesson, made from the lessons' own pictures in the course look: the picture on a card, the title big, the outcome in gold, the author small between rules. The upright 9:16 course cover fills the frame.
Say it: “Make the course with covers, by 'Jane Doe, Nurse Educator', outcome 'Handle a kitchen fire safely'.”
request
POST /v1/direct-course
{
  "brief": "A short course on kitchen fire safety for new cooks",
  "title": "Kitchen fire safety",
  "outcome": "Handle a kitchen fire safely",
  "author": "Jane Doe, Nurse Educator"
}
New in 3.69.0. Covers are on by default; cover: false leaves them out. Words: course title / outcome (else the first sentence of description) / author; per node title and outcome (else the part title, a reading page's title, the first question, or the node id said like words). A cover with no author shows the course title as its byline. Read them on GET /v1/interactive/{id} under covers { course, lessons, notMade } (notMade gives the reason for any that could not be made). Files: GET /play/{id}/covers/course.jpg and /covers/{node}.jpg. The SCORM package carries them under covers/; the LTI deep-link reply carries the course cover as the thumbnail. The gold outcome line takes up to 60 characters on two lines.
Course trailer (20-40 second film from the lessons) new in 3.69.0
Renders a vertical film from the lesson titles and outcomes: the cover, the course promise, one scene per lesson, then the cover again with a thank-you chime. Each lesson scene holds a block (lesson title small above its outcome big; a band at the top counts '1 of 5'), held as long as its words take to read.
Say it: “Make a 30-second trailer for this course.”
request
POST /v1/interactive/{interactiveId}/trailer
{
  "seconds": 30
}
New in 3.69.0. Body (all optional): preset, seconds (20-40), label, webhook. A scene is held 3.6 to 6.2 seconds; a lesson with no picture borrows the nearest lesson's (recorded in picked.notes). The reply and the course's trailer say which lessons were used and why (picked). Off unless asked.
Quiz sounds and completion confetti new in 3.69.0
A soft ping on a right answer, a quiet tick on a wrong one, and a burst of confetti on a passed lesson, using the same sounds and burst pieces as films.
Say it: “Turn on quiz sounds for this course.”
request
PATCH /v1/interactive/{interactiveId}
{
  "sounds": true
}
New in 3.69.0. sounds: true can also be sent when the course is made. A muted lesson stays silent; an exam (graded) never says right or wrong so it stays silent; reduced-motion devices get no burst; a '♪' button lets the learner switch the sounds off (remembered on that device). Off unless asked.
Learner link, results and certificates
A public player link learners open with no sign-in, with scores, results and a certificate made out to them.
Say it: “Give me the learner link and how many people finished.”
request
GET /v1/interactive/{interactiveId}/results
Learners play at https://studio.futurizenow.co/play/{interactiveId} (add ?learner=<name> for the certificate). GET /v1/interactive lists lessons; GET /v1/results is the cross-course roll-up; DELETE /v1/interactive/{id} removes one.
LMS packages (SCORM, xAPI, cmi5)
A ZIP any learning system imports.
Say it: “Give me a SCORM 2004 package for Blackboard.”
request
GET /v1/interactive/{interactiveId}/export/scorm2004
Formats: scorm (1.2), scorm2004, xapi, cmi5. SCORM embeds all media; xAPI and cmi5 are small launchers for the hosted lesson. Suggested: Moodle, Canvas, 360Learning, Totara use scorm; Blackboard, D2L, Docebo, Cornerstone, Workday, Absorb, LearnUpon, Litmos use scorm2004; TalentLMS uses cmi5. Guide: https://studio.futurizenow.co/use/lms-video-api
LTI 1.3 inside Canvas, Moodle, Blackboard, D2L
Teachers add a Studio lesson inside a course; learners open it with no second sign-in; scores go to the gradebook.
Say it: “Connect our Moodle so teachers can add Studio lessons.”
request
POST /v1/lti/registration-links
One-click link (7 days, single use). Also GET /v1/lti, GET and POST /v1/lti/platforms { issuer, clientId, authUrl, tokenUrl, jwksUrl, deploymentIds }, DELETE /v1/lti/platforms/{id}. Canvas config https://studio.futurizenow.co/lti/canvas.json; login /lti/login; launch /lti/launch; keys /lti/jwks. Tested against a stand-in; a real LMS test is still owed.

8. Covers and thumbnails

Cover or thumbnail for any film, picture or course new in 3.69.0
Result first and big, name small: the face fills the frame, the foot fades to near-black, a serif headline, a gold second line, a small-caps byline between gold rules, an optional badge.
Say it: “Make a vertical cover from my finished film: 'Number 2 on Amazon', 'in 30 days', by 'Jane Doe, Author', badge 'New'.”
request
POST /v1/cover
{
  "from": "<finished film id>",
  "headline": "Number 2 on Amazon",
  "accent": "in 30 days",
  "byline": "Jane Doe, Author",
  "badge": "New",
  "size": "9:16"
}
New in 3.69.0. Source: from (one of your films) or source (a picture or video: URL, data-URL, stored asset). headline up to 90 characters (required). size: 9:16 (default 1080x1920), 1:1, 16:9. look: testimonial | course | minimal (left out: a person gets testimonial, a picture or diagram gets course). at (seconds) picks the frame, otherwise Studio scores up to twelve and takes the best. brand { color, accent }, format jpg | png, vision: false keeps the picture on the engine. Poll GET /jobs/{id}: result.coverUrl and result.picks (frame, look, crop, type size, notes), so any choice can be overridden. Not billed as a film.

9. Social and export

Schedule finished films to social networks
Post to LinkedIn, Instagram, Facebook, X and YouTube from the maker's own connected accounts, through FN Pulse. Studio checks each film against each network's rules before booking.
Say it: “Post this film to Instagram and LinkedIn on Friday at 9.”
request
POST /v1/social/schedule
{
  "items": [
    {
      "film": {
        "job": "<jobId>"
      },
      "platforms": [
        "instagram",
        "linkedin"
      ],
      "at": "2026-12-04T09:00:00Z",
      "text": "Our new film.",
      "title": "New film"
    }
  ]
}
Studio side live since v3.57.0; whether the Pulse side and box settings are fully set up is UNCONFIRMED, and these routes are documented for the signed-in Studio (Bearer-key use not confirmed). Also GET /v1/social/accounts, POST /v1/social/connect/{network}, GET /v1/social/schedule, DELETE /v1/social/schedule/{id}, GET /v1/social/film?job=, PUT and DELETE /v1/social/own-key. Per-network captions via texts { instagram: '...' }; youtubeTitle (100 characters).
Download, webhooks and delivery checks
Files for sharing, completion callbacks, and an automatic quality check on every finished film.
Say it: “Tell my server when it's done, and warn me if it is too big for WhatsApp.”
request
POST /v1/direct
{
  "brief": "A 20-second welcome film",
  "webhook": {
    "url": "https://your-server.example/studio-hook"
  }
}
Downloads: GET /files/{id}.mp4, .srt, .vtt, .jpg (poster), .chapters.vtt (key required) or a 6-hour link from POST /v1/download-link. Webhook is signed and retried; X-Merge-Delivery stays the same across retries; treat it as a nudge and fetch GET /jobs/{id}. Job fields: advisories[], filmcheck, deliverySizes (fits telegram 50 MB, email 25 MB, whatsapp 16 MB), contactUrl, filmcheckUrl.

10. Account and API

Keys, plans, usage and your customers
Keys are made by a signed-in person in Studio, Developers; usage is split per customer; plans via Stripe.
Say it: “How many films did my customer Acme make this month?”
request
GET /v1/usage
GET /v1/keys, POST /v1/keys { name } (shown once), DELETE /v1/keys/{id} (signed-in session, not a key). GET /v1/plans, GET /v1/account, POST /v1/account/checkout { plan: creator | studio }, POST /v1/account/portal, POST /v1/account/ask { note }. Free start: three short films a month, no card. Per-customer: X-On-Behalf-Of header and GET /v1/jobs?endUser=. Errors: invalid_request 400, unauthorized 401, plan_limit 402, forbidden 403, not_found 404, conflict 409, too_many_requests 429, internal_error / provider_failed / unavailable 5xx.
Health, version and live lists
Probe what this Studio can do right now; build pickers from the lists.
Say it: “What version is Studio and what's switched on?”
request
GET /health
No key needed. Lists (cache daily): GET /v1/themes, /v1/presets, /v1/effects, /v1/atmosphere, /v1/meshes, /v1/voices, /v1/cast, /v1/score, /v1/music, /v1/sfx (new in 3.69.0), /v1/templates, /v1/library, /v1/draw. GET /v1 is the full live index with your key.
Not confirmed: 3D talking characters (built in the engine, no request exposes them yet); social scheduling through an API key and the posting service's own setup; pen-mark find (needs the picture-reading service); a real learning-system test of LTI 1.3; the licensed music pack ships empty.

✦ What's new

Recent capabilities you can call

The engine self-describes at runtime — GET /health reports the live version and which capabilities are on. These are the notable recent additions:

New in 3.69.0: courses share the story look, and Indic captions are shaped correctly
Released in 3.69.0; GET /health reports the version. Every ready course gets a cover for the course and one per lesson (cover: false opts out; read them under covers on GET /v1/interactive/{id}). POST /v1/interactive/{id}/trailer renders a 20-40 second vertical trailer from the lessons. sounds: true adds a ping on a right answer, a tick on a wrong one and confetti on a passed lesson. A scene can carry hand-written blocks with scene.frame.phrases. Captions burned in Hindi, Marathi, Bengali and the other Indic scripts now join letters and place vowel signs correctly, and covers typeset those scripts too. director: "auto" chooses marks and effects from what is said. See What Studio can do.
Characters from the creative library
Ask for a character in plain words — POST /v1/library { "request": "a tired night-shift nurse", "kind": "character" } — and it comes from the shared creative library, or is made on the spot as a small recipe (age, build, skin, hair, clothes, gender — no voice of its own) with a preview portrait. Put it in any cast as { "id": "nurse", "library": "lib_…" }; its voice defaults to a gender-matched voice from the voice library (set voice on the entry to choose another; list at GET /v1/voices); other fields on the entry change it for your film ("top": { "color": "#E53935" }). Only generic characters are kept for everyone (a nurse, a farmer, a firefighter); a real person, a brand, a film character or your customer's own is made for that film alone, as an original person, and never kept. No photo and no voice sample is involved — a real person's avatar is /v1/people, which is separate. Full details in llms-full.txt.
Every film gets a score — synthesised, not licensed
A film made from a brief used to play voice-on-silence: measured on a real 70s Director film, 18.1% of it was absolute silence, including the first 4.8 seconds. POST /v1/direct now writes the film a music bed and mixes it under the narration, ducking on every spoken line. It is generated by the engine — there is nothing to license, nobody to attribute and no vendor that can go down. Choose one with "score": "warm" | "calm" | "bright" | "serious" | "tender", decline it with "score": false, or send your own music — yours always wins. List them at GET /v1/score; the finished job reports film.score.
Fields we don't understand are named back to you
A misspelled parameter used to be dropped in silence — targetSeconds on /v1/direct was ignored, the default length took its place, and the finished film reported a target you never set. Every submit now answers with ignored: [{ field, didYouMean }] and a plain-language warning, and the same report rides the job as a field-not-understood advisory. Nothing is refused — a field we haven't heard of never breaks your request, it just stops being invisible.
Say how long the film should be, however you spell it
/v1/direct takes target: { seconds }, targetSeconds, or seconds — all the same thing. A length that isn't a number is refused at the door with the field named, rather than quietly becoming 60. Note that narrated scenes run as long as their voice takes, so a target is aimed at, not guaranteed; the job says so.
Pin the picture service, like the voice
"imageProvider": "openai" | "google" on /v1/direct and /v1/direct-course. A pinned service is the only one tried — if it refuses you're told, never handed the other one's pictures in a different house style. Worth knowing: only Imagen (google) draws to a 16:9 frame; gpt-image-1's largest landscape is 1536×1024, so a 1080p film enlarges it slightly and eases the Ken Burns move to keep that from showing.
Scene headings that place themselves clear of the subject
Opt-in: put "heading": "The referral engine" on any scene (or { text, position } to pin it), or send "headings": true to /v1/direct / /v1/direct-course and the Director labels each scene itself. Left to place itself, the engine measures where the picture is quiet — edge energy per zone, because a face or a subject is where the detail is — and files the label top, left, or right accordingly, never over the busy part and never in the caption band. Films that don't ask render byte-identically.
Re-produce a course without re-buying the parts that didn't change
Send "previous": "<interactiveId>" on POST /v1/direct-course when producing a new version of a course you already made. Parts whose teaching the edited brief leaves untouched keep their finished films — no new pictures, no new narration, no render, nothing metered — and each kept part says so on its own timeline (part-reused). Anything the edit touches is made fresh, and when in doubt the engine rebuilds: a stale film can never wear an edited script. Reuse needs the same course-wide settings (theme, voice, size, language, length, tone, rules, providers) — change any of those and every part rebuilds, with a note saying why. Courses produced before this feature carry no reuse record; their first re-produce is full price and starts the chain.
Locked rules — your non-negotiables, on every variation
PUT /v1/rules stores your standing rules ("no music", "never show faces") server-side, per key. They are merged into every generation — reels, Director films, courses, storyboard images, redraws and re-runs — so a rule stated once can never be forgotten by a later variation. Read back with GET /v1/rules.
Your words, verbatim — captions that never mishear
Pass lyricsText to POST /v1/reel or POST /v1/caption and the captions carry your exact text, aligned word-by-word to the audio's clock. Transcription only provides timings; a misheard name or a skipped line can no longer reach the film. The job notes report how much of the text anchored to the audio.
Silence is reported, not just left blank
A transcriber fed room tone does not answer with silence — it answers with a sentence it has heard before ("Thank you for watching."), with a start and an end. The engine refuses to burn those into anyone's film, and now says so: an invented-words-over-silence-were-dropped advisory on the job carries every such stretch with its clock (dropped: [{ start, end, text }]), so "no captions appeared" reads as "nothing was heard between 0:03 and 0:08". A recording that was room tone end to end fails with those stretches named. Applies to POST /v1/caption and POST /v1/enrich alike.
Take the music out — vocals kept
POST /v1/isolate takes any audio or video (media: URL, data-URL or stored asset) and returns the voice alone — music, instruments and noise stripped. A video comes back as the same video (picture stream-copied) with only the voice on its soundtrack. Poll /jobs/{id}; the outputs land in result.vocalsUrl / result.videoUrl.
Big media rides the asset store everywhere
POST /v1/caption, /v1/enrich, /v1/reel and /v1/isolate now accept a stored-asset reference ({"__asset": "…", "mime": "…"} from POST /v1/assets) as the video/song — so a 500MB film streams up once instead of failing the 32MB JSON cap. Transcription is fed a small mono extract, never the whole file.
Durable everything (Cloudflare R2)
With R2 configured (GET /health → "storage": "r2"), the box's disk is a CACHE, not a home: uploads, recordings, generated stills, finished films, interactive course media, certificates, learner results, templates, locked rules and the usage ledger all persist to object storage. A swept file restores transparently on next access, and a brand-new box pulls its state home from the bucket at boot.
Record on your phone → lands on your computer
The Recorder shows a QR; scan it with any phone (no sign-in — the token is the capability), record, and the clip uploads straight into the exact slot on the computer that made the QR, while a copy saves to the phone. API: POST /v1/handoff (mint), GET /v1/handoff/{token} (poll), POST /v1/handoff/{token}/asset (public phone upload).
Record in the browser (Studio Recorder)
A new page at /studio/record records the webcam up to 4K with native noise-cancellation, a WebGL chroma-key greenscreen, native background blur, and AI background blur / replacement (MediaPipe body-segmentation, vendored and served by the engine — no external request) — then auto-saves the clip as one of your sources. Reachable from the Studio's 🎥 Record button.
Upload a recording as a source
POST /v1/assets streams a browser-recorded clip (4K, camera permitting) — or any large media — straight into the content-addressed store and hands back a {__asset} marker you can drop in any timeline, plus a Range-served preview URL. The backbone of the in-Studio camera recorder: record in the browser, and the clip becomes another of your video sources. See section 02.
Exam-grade scoring (server-graded mode)
Set "graded": true on an interactive and the answer key never reaches the browser — the player submits each response to POST /play/{id}/grade, the engine decides correctness, and the score is server-authoritative (a forged client score can't earn a certificate). Default courses stay client-graded for fast, offline learning checks. See section 04.
Full interaction set: 8 question types
Interactive courses now take dragdrop, match, ordering (arrange in order), fillblanks (cloze), markwords (tap the right words), and hotspot (click a region of an image) — alongside choice/multi/text. All work by mouse, touch, or keyboard, score automatically (server-graded when the course is graded), and report through SCORM/xAPI/cmi5. See section 04.
Your key is a tenant boundary
Your named key now sees, downloads, and cancels only the jobs and batches it submitted — GET /v1/jobs, /v1/batches, and /v1/usage return your own work; another product's ids, files, and spend are invisible (a request for one answers 404).
Soft cinematic transitions, at speed
Feathered turns — breathe, veil, curtain, bloom — now render several times faster at identical quality. Use them on any scene junction.
Weather & light atmosphere
Per-scene atmosphere — snow, embers, fog, light — layered over your picture. List them at GET /v1/atmosphere.
Branching interactive courses + LMS export
Build a graph of nodes with scored questions and remediation loops, then export SCORM 1.2/2004, xAPI, or cmi5 — see sections 04 and 05.
Captions & translation
Burn-in captions, or auto-transcribe a supplied clip; subtitle translation across many languages (GET /health → captionLangs).
HeyGen presenter avatars (opt-in)
If you'd rather we generate a talking-avatar than have you supply the clip, a key-gated presenter path exists. Ask the operator to enable it; then pass presenter: true with your avatar/voice ids.

01 Authentication

One bearer key, held server-side

Every call except /health and the public learner player carries your token:

http
Authorization: Bearer YOUR_API_KEY

Your app is issued its own named key, so usage is attributed to you and can be revoked independently — never share the master key, and never ship the key to a browser. Every job records which key submitted it, and GET /v1/usage returns your per-key metering.

Your key is a tenant boundary: it sees, downloads, and cancels only the jobs and batches it submitted. Another product's ids, files, and spend are invisible to you, and a request for one answers 404. (The operator's console key and admins see everything, for support.)

02 Providing content

Pass your assets by URL

Anywhere a visual takes a src, you may pass:

FormUseNotes
https://…Recommended for HeyGen / Higgsfield clips & imagesStreamed to disk, size-capped; host must be publicly reachable
data:<mime>;base64,…Small inline assets onlyNever inline hundreds-of-MB video — use a URL

Accepted visual kinds: video, image, color (a themed card). Private, loopback, and cloud-metadata hosts are refused by the SSRF guard — public URLs are fine. The request body itself is capped at 32 MB, the other reason to pass big media by URL.

03 Stitch content into a video

POST a timeline, poll the job, fetch the film
POST/v1/render→ { jobId }

Each scene carries a duration, a visual, and optional transition, overlays, effects, and atmosphere. Your provided clip or image is the scene's visual.

json — request body
{
  "timeline": {
    "preset": "16:9",
    "theme": "lux",
    "scenes": [
      {
        "duration": 6,
        "visual": { "type": "video", "src": "https://your-cdn/heygen-intro.mp4" },
        "transition": "fade",
        "overlays": [{ "type": "title", "text": "Welcome" }]
      },
      {
        "duration": 5,
        "visual": { "type": "image", "src": "https://your-cdn/slide-1.png", "motion": "in" },
        "overlays": [
          { "type": "eyebrow", "text": "MODULE 1" },
          { "type": "title",   "text": "Fire safety basics" }
        ]
      }
    ]
  }
}
FieldValues
preset16:9/landscape/course, 9:16/portrait/vertical, 1:1/square. Omit → 16:9.
visual.typevideo (your clip), image (still, optional Ken-Burns motion: in / out / drift-left / drift-right / punch), color
overlays[].typeeyebrow, title, subtitle, lowerthird, keyline, caption, popcaption, lockup
transitionfade, cut, breathe, veil, curtain, bloom, pixelize, whip … (live: GET /v1/effects)
captions{ "subtitles": [{ "start", "end", "text" }] } for burn-in captions, or omit. Burned into the picture and shipped beside it as .srt + .vtt
languagea BCP-47 tag (fr, pt-BR) — reaches the .vtt header and the mp4's own track tags
bookendthe film fades up from black and out to it, and a music bed comes up rather than starting at full level. false keeps the audio ramps and drops the picture fades — for clips played back to back

Poll the job

GET/jobs/{jobId}
json — while it works
{ "status": "processing", "progress": "scene 3/12 built",
  "phase": "building", "percent": 25, "step": { "done": 3, "total": 12 } }

progress is a sentence to show a person — don’t run a regex over it. Beside it, phase is the stable key (GET /v1 lists them in order), percent is how far through that phase, and it is absent rather than stale when there is no honest number.

json — on done
{ "status": "done", "duration": 11.0, "seconds": 11, "label": "Fire safety",
  "url": "https://studio.futurizenow.co/files/{id}.mp4",
  "captionsUrl": "…/files/{id}.srt", "subtitlesUrl": "…/files/{id}.vtt", "posterUrl": "…/files/{id}.jpg",
  "chaptersUrl": "…/files/{id}.chapters.vtt",
  "createdAtIso": "2026-08-06T17:41:58.887Z", "restartUrl": "…/jobs/{id}/restart",
  "film": { "width": 1920, "height": 1080, "fps": 30, "bytes": 8431276, "scenes": 4,
            "videoCodec": "h264", "audio": "stereo 48 kHz", "captionCues": 23,
            "chapters": [ { "index": 1, "start": 0, "end": 3.3, "title": "Fire safety basics" } ] } }

status runs queued → processing → done | failed. Every URL in every answer is whole — nothing here needs joining to a base. The film comes back described: its size, its codecs, how many scenes, and its own table of contents (written into the mp4 as well). Fetch the outputs at GET /files/{id}.mp4 (Range/seek supported, plus .srt, .vtt and .jpg); they carry an ETag, so a re-check costs a 304.

Prefer streaming over polling. GET /jobs/{jobId}/events is an SSE stream of named events — progress, then a final done/failed — each frame a whole job view. Reconnect with Last-Event-ID and the numbering carries on where it left off; you are handed the current state, never a replay of stale ones. Or add "webhook": { "url": "https://your-app/hook" } to the POST and we call you on completion, with the same links, whole.

04 Stitch content into a course

POST a node graph, get a learner player URL
POST/v1/interactive→ { interactiveId, playerUrl }

Each node is an ordinary timeline (same schema as above) that renders to its own clip. A node ends in a question whose choices route to other nodes, a next that auto-advances, or nothing (the end). Loops are legal — a wrong answer can route to a remediation node that returns to the question.

json — request body
{
  "title": "Fire Safety Module",
  "start": "intro",
  "theme": "lux",
  "nodes": {
    "intro": {
      "timeline": { "scenes": [
        { "duration": 6, "visual": { "type": "video", "src": "https://your-cdn/heygen-lesson.mp4" },
          "overlays": [{ "type": "title", "text": "Lesson 1" }] }
      ]},
      "question": {
        "prompt": "A pan catches fire. What do you do?",
        "choices": [
          { "label": "Cover it with a lid", "next": "correct", "correct": true },
          { "label": "Throw water on it",  "next": "recap" }
        ]
      }
    },
    "recap":   { "timeline": { "scenes": [ { "duration": 5, "visual": { "type": "color", "src": "#301414" }, "overlays": [ { "type": "title", "text": "Not quite" } ] } ] }, "next": "intro" },
    "correct": { "timeline": { "scenes": [ { "duration": 4, "visual": { "type": "color", "src": "#12301c" }, "overlays": [ { "type": "title", "text": "Well done" } ] } ] } }
  }
}

Poll GET /v1/interactive/{interactiveId} until every node has rendered. The learner-facing player is public at GET /play/{interactiveId} — the unguessable id is the capability token; hand that URL to learners or embed it. Results and analytics come from GET /v1/interactive/{interactiveId}/results. Optional per node: a certificate minted on pass, a resultsWebhook, and templateName + row to mail-merge inside a branch.

05 LMS export

Package a rendered course for any LMS
GET/v1/interactive/{id}/export/scormSCORM 1.2 · zip
GET/v1/interactive/{id}/export/scorm2004SCORM 2004 · zip
GET/v1/interactive/{id}/export/xapixAPI / TinCan · zip
GET/v1/interactive/{id}/export/cmi5cmi5 · zip
SCORM 1.2 / 2004 embed the media — a portable, offline-capable package (larger). xAPI / cmi5 are thin launchers that point at the hosted player. Use SCORM when the package must be self-contained; use xAPI/cmi5 when your LMS can reach the player URL.

06 Discovery

Build against live catalogs, don't hardcode
EndpointReturns
GET /healthcapabilities & version (no auth)
GET /v1/themesvisual themes
GET /v1/presetsformat presets (16:9 / 9:16 / 1:1)
GET /v1/effectscinematic looks & transitions
GET /v1/atmosphereweather / light layers
GET /v1/scoremusic beds the engine writes
GET /v1/voicesspeech services & their voices
GET /v1/usageyour per-key metering
GET /v1/librarythe shared creative library: backgrounds, pictures, sets, props & characters

07 Notes & limits

A few things to design around
  • Async by design. Renders take seconds to minutes — never block on the POST. Poll /jobs/{id}, subscribe to /jobs/{id}/events, or use a webhook.
  • Caps. A single render takes up to 500 scenes; an interactive graph up to 40 nodes. Ask if you need these raised.
  • Cancellation. DELETE /jobs/{id} cancels a job (and stops any in-flight spend); DELETE /v1/interactive/{id} cancels a course.
  • Provide, don't generate. When you supply the assets, no provider keys are involved. The engine can also generate — AI Director, stills, narration, HeyGen presenter — but that's a separate, key-gated path.

08 Quickstart

One render, end to end
bash
# submit
JOB=$(curl -s -X POST https://studio.futurizenow.co/v1/render \
  -H "Authorization: Bearer $KEY" -H "content-type: application/json" \
  -d '{"timeline":{"preset":"16:9","scenes":[
        {"duration":6,"visual":{"type":"video","src":"https://your-cdn/clip.mp4"},"overlays":[{"type":"title","text":"Welcome"}]},
        {"duration":5,"visual":{"type":"image","src":"https://your-cdn/slide.png","motion":"in"}}
      ]}}' | jq -r .jobId)

# poll
curl -s https://studio.futurizenow.co/jobs/$JOB -H "Authorization: Bearer $KEY" | jq .status

# fetch when done
curl -s https://studio.futurizenow.co/files/$JOB.mp4 -H "Authorization: Bearer $KEY" -o film.mp4