Can I vibe code Descript?
Edit video and podcasts by editing the transcript text.
Edit-the-text-edit-the-video is deep infrastructure wearing a friendly UI: word-level forced alignment, gapless waveform re-rendering, filler-word detection, studio sound, and voice cloning. You can vibe-code a transcript viewer that jumps to timestamps in an afternoon, but cutting media by deleting words — reliably, at word boundaries, without artifacts — is a research-grade pipeline, not a weekend build.
Build a transcript-first audio review tool (a lightweight cousin of Descript, not a clone). Let me upload an audio file, transcribe it with the OpenAI audio API with word-level timestamps, and render the transcript with each word clickable to seek the player. Let me select a text range and export that audio segment (use ffmpeg via a server route). Store projects in a database with the audio in object storage. Make the transcript editable for correcting words (text only — no audio editing claims). Clean Next.js app, commented, deployable to Vercel.
A transcript-alongside-player tool: upload audio, get a Whisper transcript with clickable timestamps, and export selected ranges — useful, but not text-based editing.
- True text-based cutting with clean word-boundary edits
- Overdub/voice correction and studio-sound enhancement
- Multitrack timeline, scenes, and screen-recording integration
- Filler-word removal that doesn't leave audible seams
The alignment-and-render pipeline is years of specialized work; every alternative that looks close collapses on edge cases the moment you cut mid-sentence.
A prompt is the first move, not the whole game. Course 01 teaches you to take a prompt like this one to a shipped, working app — reviewing, correcting, and steering the agent the whole way.
Start Course 01 →