AutoVideoStudio

How AutoVideoStudio works

1. Brief

Topic, format (documentary, Top-10, explainer, story), length 1–60 min, language, aspect ratio.

2. Research

We read the top Wikipedia articles for the topic (CC BY-SA). Each becomes a numbered source S1, S2…

3. Script

Our own AI writes the script on Surge servers; every scene lists the sources it uses. If AI is unavailable, an extractive script is built from source sentences.

4. Fact-check gate

Every sentence with a number or a name is matched to its source. Wrong numbers or unsupported claims block the render until you edit or approve them.

5. Voiceover

Sarvam Bulbul v3 voices, your consented voice clone, or upload your own recorded narration.

6. Footage

Pexels, Storyblocks, Pixabay video first, then photos and Wikimedia Commons (public domain / CC0 / CC BY / CC BY-SA only). Every file goes into the asset ledger.

7. Edit

Retention plan: fast cuts in the first 30 s, pattern interrupts at least every 20 s, Ken Burns motion on stills, captions, music ducked under the voice, loudness −14 LUFS.

8. Publish

Two thumbnails (A/B), title, description with chapters, sources and credits, tags. Upload to YouTube as private or scheduled.

9. Shorts & dubs

Best 15–58 s stretches re-cut vertical; full video dubbed into other languages when multi-language is switched on.

10. Scene editor & chat edits

Every shot shows a relevance score against its narration. Swap a shot, re-search a scene, generate an AI image, upload your own, or type the change ("caption style karaoke", "swap visuals in scene 2: monsoon clouds").

API

Studio plan. Authenticate with a workspace API key (Settings → API key).

POST /api/v1/videos
Authorization: Bearer svs_live_...
{ "topic": "History of the Indian Railways", "format": "documentary",
  "targetMinutes": 8, "language": "te-IN", "shorts": 2, "dubLanguages": ["hi-IN"] }

→ 201 { "id": "prj_...", "status": "queued" }

GET /api/v1/videos/prj_...
→ { "status": "done", "video": "/api/files/prj_.../video.mp4",
    "costPerFinishedMinute": 0.03, "factcheck": { "supported": 41, "total": 41 } }

Copyright & licences

Each project has an asset ledger (provider, licence, licence URL, attribution, page URL, SHA-256 of the exact file). CC BY / CC BY-SA credits are written into the YouTube description automatically. Non-commercial and no-derivatives licences are rejected. Music is generated, so it never triggers Content ID.