transcriptcut

Appearance

transcriptcutEdit video by editing its transcript

Built for video podcasts. Delete a sentence in the transcript and the matching cut lands on the timeline, then get podcast-ready audio, speaker labels, chapters, show notes and Shorts from the same project. Local-first and open source: no cloud account, just your own OpenAI API key.

Download for macOS
Get Started
Podcast Workflow
See Screenshots
View on GitHub
transcriptcut editor showing video, transcript, and timeline
📝

Transcript-first editing

Select and delete text in the transcript and the corresponding section of the video is cut — reflected on the timeline and skipped during playback, with full undo/redo.

🎙️

Made for long, multi-speaker episodes

Multi-GB uploads, background transcription in chunks with progress, a 720p editing proxy, and speaker detection with one-click rename and "cut everything this person said".

🔊

Podcast-ready audio

Hit -16 or -14 LUFS, reduce noise, level a quiet guest against a loud host, remove rumble. Preview 15 seconds before and after, then bake it into the export.

📚

Chapters, show notes, MP3

AI chapters built into every MP4 and MP3 export and copyable as YouTube timestamps, editable show notes and title ideas, TXT/Markdown transcripts, and MP3/WAV export for your feed.

📱

Shorts from the same edit

AI finds the episode's best standalone moments. Render them as 9:16 clips with big word-by-word captions and a crop you position on the video.

🎯

Scoped AI actions, not a chat bar

Every AI step is a specific button: filler words, speakers, chapters, show notes, highlights. Models return text and sentence picks, never timestamps or edits. No open-ended "ask the AI to edit this" field — see why in the Guide.

🖥️

Runs entirely on your machine

SQLite via Prisma, local filesystem storage, FFmpeg as a child process. The only external calls are to OpenAI, for transcription and the AI buttons — never for storage.

🎞️

Zoomable timeline + waveform

Frame thumbnails and a live waveform that get denser as you zoom in, not just stretched pixels.

🎨

Filters, properties, captions, elements

Color presets, manual color adjustment sliders, karaoke-style captions, and a logo/watermark overlay — all previewed live and baked into the final export.

🔒

No accounts, by design

v1 is a single local user. That's a deliberate simplicity choice for a self-hosted tool, not a gap — see the security guide before exposing it beyond your own machine.

Why transcript-first? ​

Traditional video editors make you scrub a timeline to find the part you want to cut. transcriptcut inverts that: the transcript is the editing surface. Read the words, delete the ones you don't want, and the video updates underneath you.

text
Transcript:  Hello everyone [DELETE] um basically [/DELETE] today we are learning AI.
                                  ↓
Timeline:    a cut is created at the exact word-level timestamps of the deleted span

Word-level timestamps (from Whisper) make this mapping exact, not approximate — see Editing Workflow for how it works end to end.

Filler-word detection, in the same view ​

AI tools panel with Find filler words and Find long pauses, flagged words underlined in the transcript

See more screenshots →

Released under the MIT License.

transcriptcut is local-first, open source software.