Audio & videoCAN I VIBECODE
Descript?
BUILD FIRST
- One specific customer type and one primary use case
- A vertical slice that completes the main job without manual workarounds
- Visible limits, review steps, and a fallback when an external dependency fails
LEAVE FOR LATER
- Broad third-party integrations
- High-scale collaboration, syncing, and permissions
- Automation that can cause harm without human review
RESEARCH NOTES
What this tool actually does.
Descript is an AI-powered audio/video editor that transcribes media into an editable text document; users cut, rearrange, and trim by editing words, and the underlying media updates accordingly. Official materials describe multitrack non-destructive editing, built-in recording, remote Rooms sessions, AI Speakers (text-to-speech), scenes/layouts for visual structure, stock elements, captions, Studio Sound (noise reduction/audio cleanup), Eye Contact (gaze correction), Underlord (AI co-editor chat), automatic clip generation, voice-cloning Regenerate, cloud sync, timestamped comments, publish links, 4K export, and integrations. Pricing/plan specifics were not independently verified.
RESEARCHED BUILD PROMPT
Give this to your coding agent.
Build a web app prototype called 'ScriptCraft' that reproduces the core Descript-style workflow: upload an audio or video file, get a transcript, and edit the media by editing the text. Users upload a file; the backend runs speech-to-text with word-level timestamps and returns a JSON timeline. The UI shows the transcript as a document with a wordbar/waveform strip: deleting or moving words in the document must cut or rearrange the corresponding media segments, and clicking a word seeks playback to that timestamp. Add simple play/pause and a playhead synced to the transcript. Export the edited audio as a single WAV/MP3 file (video export is out of scope). Implement with React + Vite front end and a Node/Express API route; use OpenAI Whisper API (or local whisper.cpp) for transcription, but include a mock fallback mode with sample transcript fixtures so the app runs without API keys. Out of scope: video rendering, multitrack timelines, voice cloning/regeneration, captions, collaboration comments, publishing links, templates, Rooms, and any Descript branding, logos, or UI layout. The UI must look original — a clean custom editor, not a copy of Descript's interface. Write a README covering setup, environment variables, run instructions, and clear limitations (Whisper latency, accuracy on noisy audio, no video output, mock-mode behavior).
Do not copy the original brand, proprietary assets, or UI; create an original visual system.