An AI director that storyboards before it spends your tokens
Frametell turns a sentence or a PDF into a narrated 1080p explainer video — with you approving the storyboard first.

What it does
Frametell is a desktop app (Electron) that turns an idea, notes, PDFs, slides or images into a finished explainer video: storyboard, narration script, voiceover, illustrated pages, and a 1080p MP4 with subtitles. A director agent drafts the storyboard and waits for your sign-off before drawing anything. It works with Anthropic, OpenAI, Google, DeepSeek, OpenRouter, or any OpenAI/Anthropic-compatible endpoint, and keeps your materials and keys on your machine.
The interesting bit
Instead of generating video from scratch, the agent drives a structured video engine with styles, components and camera moves — which the README credits for stable results and fewer tokens. And when the output is 95% right, you fix the last 5% by hand: finished films open in a click-to-edit editor where hand edits are a layer over the agent’s work, cost zero tokens, and every save is a version in the film’s own git history.
Key highlights
- Storyboard approval gate: nothing is drawn until you sign off, and you can point at a page and ask for changes
- Live cost and time display at every step — the demo film (72 s) cost $1.59 and took about 13 minutes
- Five built-in styles (
warm-story,study,chalkboard,whiteboard,tech-dark) - Voiceover via ElevenLabs, or Volcengine TTS for Chinese; pages are self-checked with lint and screenshots
- Runs fully locally apart from the model APIs; needs Chrome/Edge/Chromium for drawing and rendering
Caveats
- The packaged app is macOS Apple Silicon only; Windows and Intel Mac users must run from source
- AGPL-3.0 licensed — network-hosted modified versions must share source (commercial licenses available)
- Still needs paid API keys for both a model and voiceover, so “local-first” means local data, not local compute
Verdict
Worth a look if you explain things for a living and would rather not learn a video editor. If you wanted fully offline, zero-cost generation, this isn’t that — the models still bill you.
Frequently asked
- What is sitor-ai/frametell?
- Frametell turns a sentence or a PDF into a narrated 1080p explainer video — with you approving the storyboard first.
- Is frametell open source?
- Yes — sitor-ai/frametell is open source, released under the AGPL-3.0 license.
- What language is frametell written in?
- sitor-ai/frametell is primarily written in TypeScript.
- How popular is frametell?
- sitor-ai/frametell has 527 stars on GitHub.
- Where can I find frametell?
- sitor-ai/frametell is on GitHub at https://github.com/sitor-ai/frametell.