Based on current runtime
Each stage exists; the automatic handoff does not
SummaryBot accepts a YouTube link and summarizes the extracted transcript; HumanBot edits separately pasted text under a fact-and-meaning preservation contract; VoiceBot synthesizes audio from separately sent text. No runtime contract promises an automatic handoff or video editing.
If transcript extraction fails, the first stage cannot complete.
The boundary of each stage
| Stage | Input | Output | Human check |
|---|---|---|---|
| SummaryBot | Public link | Notes and timestamps | Matches the video |
| HumanBot | Your draft | Edited text | Facts, tone, attribution |
| VoiceBot | Final text | Audio file | Pronunciation and rights |
Practical workflow
- 1. Create the notes: Send a public link with an available transcript to @vustSummaryBot. Check the thesis, key points, and timestamps against the original video.
- 2. Build the script: Copy the useful points into your draft, then add transitions, facts, and your own position. For less formulaic wording, send that draft separately to @vustHumanBot.
- 3. Review rights and facts: Do not publish a close retelling of someone else's video as your own. Remove unsupported details, add attribution, and confirm that you may create the derivative work.
- 4. Generate the audio: Send the final text as a separate message to @vustVoiceBot. It creates an audio file from text; it does not edit the source video or receive the script from SummaryBot automatically.
Working artifact: thesis → 3–5 verified points → transitions → source attribution → final voiceover text. This is an output template, not a measured benchmark.
- Based on
- Three verified runtime contracts, with no speed or accuracy metric.
- Best for
- Lecture notes, your own videos, and permitted content repurposing.
- Not ideal for
- Exact transcription, video editing, or republishing another creator's work without rights.
- In short
- Create notes, manually build and review the script, then generate the audio.