A RECORDING BOOTH AND A TRANSCRIPTION ENGINE

Two small tools in one app. Record: read your script one passage at a time in a CRT-styled teleprompter, stack takes until one lands, and export a clean voice track. Transcribe: paste a video URL or pick a local audio/video file and get a transcript back — imported captions when they exist, otherwise on-device Whisper. Crash-safe, keyboard-first, fully local — no accounts, no cloud, no telemetry. Free and open source.

Download for macOS Download for Windows Source on GitHub

macOS 13+ · Apple Silicon · notarized .dmg Windows 10/11 · x64 · unsigned installer — on first run SmartScreen shows “Unknown publisher”; click More info → Run anyway v0.1.4 · Apache-2.0 OR MIT · release notes

The booth screen: a teleprompter passage above a record control bar

The script overview: chaptered passages with merge and split controls

How it works

Import a script (.md or .txt). Headings become chapters; paragraphs split into recordable passages you can merge and split.
Record passage by passage. SPACE to record and stop. Takes stack up newest-first; revert or delete any take — files are moved, never deleted, with a 5-second undo.
Edit the script inline. Click the teleprompter text, fix the wording, save — the change writes back to your source file.
Export a sample-perfect concatenated voice.wav (plus voice.mp3 when ffmpeg is installed), with prior outputs backed up first.

Built like a tape machine

Audio is captured natively in Rust at 24-bit — takes are flushed to disk continuously, so a crash mid-take is recovered on the next launch.
Sessions live in a booth/ folder next to your script and resume exactly where you left off.
Every action is both a key and a visible button. Press ? in the app for the full cheat-sheet.

Keys

SPACE	record / stop
P	play top take
R R	revert top take (double-tap)
✕ on a take card	delete that take (5 s undo)
ENTER	accept take ▸ next passage
J / K	next / previous passage
G	view transcript (merge · split)
TAB	review & export (⌘E)

On Windows, ⌘ shortcuts use Ctrl (e.g. Ctrl+E to export).

The transcription engine

The same app also turns video and audio into text — no separate tool, no upload. Open the transcribe screen and either paste a URL or pick a file.

Paste a URL from YouTube, TikTok, Instagram, or Facebook — or pick a local audio/video file.
If the source carries a real caption track, it's imported instantly (tagged as manual or auto subs). Otherwise Booth pulls the audio and transcribes it locally with Whisper large-v3-turbo — a 1.6 GB model that downloads once and then runs entirely on your machine (Metal on macOS, CPU on Windows).
Short-video platforms (TikTok / Instagram / Facebook) skip the caption probe and go straight to local transcription, since usable tracks are rare.
Export 8 formats — TXT, SRT, VTT, JSON, CSV, HTML, DOCX, PDF — with one-click copy buttons for the plain text.
Saved transcripts re-open from the Load screen. Downloaded media is transient and cleaned up — nothing leaves your machine.

Why we built it

Booth is the tool we use in the Attestrum studio to narrate our own video scripts. Attestrum is about verifiable, local-controlled provenance for AI training data — Booth carries the same posture into a small recording tool: your words, your audio, your disk.