How to use
- Paste a transcript (ideally with “Name:” labels), open a VTT/SRT file, or transcribe an English recording with Whisper.
- Review the action items, decisions and open questions; edit owners, tasks and dates.
- Optionally create an AI summary, then copy or download the Markdown notes or the action items as CSV.
Worked example
“Priya: I'll send the budget by Thursday” becomes the action item Send the budget (owner Priya, due Thursday); “Lena: Marco, can you book the venue before next Friday?” becomes Book the venue (owner Marco, due Next Friday).
Supported formats and limits
| Input | Transcript text, TXT, VTT, SRT or DOCX transcripts, Audio recordings up to 30 minutes (English) |
|---|---|
| Output | Markdown notes, Action items (CSV) |
| Limits | Action items, decisions and key-sentence summaries are found on your device instantly. Optional models download once from Hugging Face: SmolLM2-360M-Instruct for an AI summary (about 390 MB, or 276 MB with WebGPU; it reads the first 1,200 words) and Whisper base.en for recordings (about 78 MB). Nothing is uploaded. |
| Engine | Rule-based extraction (speaker labels, “I’ll …”, “Name, can you …”, “Action item: …”, due-date phrases, “we decided/agreed”) and TextRank key sentences; optional AI summary by SmolLM2-360M-Instruct (Apache-2.0) running in your browser with transformers.js in a Web Worker (WebGPU when available, otherwise WebAssembly on the CPU); optional Whisper base.en (Apache-2.0) transcription via transformers.js |
Limitations
- Fixed patterns miss unusual phrasing, and owners can be wrong when speakers are not labelled (Whisper does not label speakers).
- The AI summary is machine-generated by a small model and can misstate what was said.
- English only.
Questions
How are action items found?
By fixed patterns, not by AI: phrases such as "I'll ...", "Name, can you ..." and "Action item: ...", plus due-date words. "Priya: I'll send the budget by Thursday" becomes Send the budget, owner Priya, due Thursday. Unusual phrasing is missed, so review and edit the list.
What do the optional models do?
Whisper base.en (about 78 MB) transcribes English recordings up to 30 minutes, without speaker labels. SmolLM2-360M-Instruct (about 390 MB, or 276 MB with WebGPU) writes an AI summary from the first 1,200 words. Both are downloaded once from Hugging Face, and their output is machine-generated and can misstate what was said.
Is the meeting uploaded?
No. Transcripts, recordings and notes stay in this tab. Only the optional model files are fetched from Hugging Face.
Privacy
On-device model. Processing runs in this browser. The open-source model files are downloaded once from the model host (Hugging Face) and cached; your content is not uploaded.
- Hugging Face: Only after you press Download model: requests for the model files (only for the optional AI summary (SmolLM2-360M-Instruct, about 390 MB) and recording transcription (Whisper base.en, about 78 MB)) from huggingface.co and its CDN, which see your IP address. Your text and files are never sent.
See the privacy policy for how toolsdocks handles data.