On-device AI workspace for Mac audio and video creators
面向 Mac 音视频创作者的端侧 AI 工作台
Website · Skills · Features · Voice & TTS · Video Demo · Download
Cliptate is a native macOS app for turning speech into editable text, subtitles, and rough cuts. Transcribe recordings, dictate into other apps, trim pauses, and generate speech with models running on your Apple Silicon Mac.
GitHub downloads are free: all models and Pro features are unlocked, with no subscription or purchase required. Optional online AI services use your own provider configuration.
Watch the Cliptate video demo on YouTube.
Light appearance with a light-gray page background.
Drop videos into the import area, or click it to select one or more files. Use the sidebar to switch tools; Settings is always available at the bottom.
The interface supports English, Simplified Chinese, Japanese, Korean, and German; choose your display language in Settings → Language.
Get started · Features · CLI · Privacy · Build from source
- Download
Cliptate.dmgfrom GitHub Releases. For daily builds, use the Nightly prerelease. - Open the DMG, drag Cliptate.app into Applications, and launch it.
- Open Models and download the models you need before your first transcription or speech synthesis session.
Release packages use Developer ID signing and Apple notarization. Each download includes a SHA-256 checksum file; see the installation and verification guide. Direct-download builds support in-app updates.
| Requirement | Details |
|---|---|
| Mac | Apple Silicon (M1 or later) |
| System | macOS 15 Sequoia or later |
| Storage and memory | Depend on the selected models; check their sizes in Models |
| Network | Needed for initial model downloads and optional online services |
- Recorded audio or video: open AI Cut, import files, select a transcription model and language, then review the transcript and pause cuts before exporting.
- Live dictation: open Voice To Text, choose a model, configure a shortcut, and grant microphone access. Direct builds use Accessibility permission to deliver text into other apps; clipboard paste is the fallback.
- Speech synthesis: select a TTS model, enter text, and choose a preset voice, reference audio, or voice description supported by that model.
| Workflow | What you can do |
|---|---|
| Offline transcription | Batch-transcribe audio and video with Parakeet, Qwen3 ASR, FireRed, or Cohere models; choose an engine suited to your language and hardware. |
| Voice dictation | Stream speech to text, use a custom vocabulary dictionary, and invoke dictation with a configurable shortcut. |
| AI Cut | Detect pauses with voice activity detection (VAD), review segments, and export a rough cut. |
| Speaker labels | Use Sortformer diarization to distinguish speakers in supported transcription pipelines. Silero VAD detects speech without identifying speakers. |
| Speech synthesis | Use Qwen3 TTS for preset voices, voice cloning, emotion/style prompts, and voice design. Available controls depend on the selected model. |
| Correction and translation | Use an optional online AI provider for transcript correction and keyword emphasis; translate subtitles with Apple Translation for supported language pairs. |
| Command line | Transcribe files, generate WAV speech, and manage custom vocabulary from scripts or AI agents. |
| Output | Formats and use |
|---|---|
| Editing timeline | FCPXML for Final Cut Pro |
| Subtitles and transcripts | SRT, styled ASS, and timestamped TXT from the app |
| Video | MP4, MOV, M4V, with optional burned-in subtitles |
| Audio | M4A, WAV |
| CLI transcription | Text, JSON, Markdown, SRT, VTT |
Use subtitles and rendered media in editors such as DaVinci Resolve or Adobe Premiere Pro, subject to each editor's format support. FCPXML is the Final Cut Pro timeline workflow.
The CLI is included in direct-download builds. Move the app to Applications first, then open Settings → Command Line and run the copied installation command in Terminal. Add ~/.local/bin to your shell's PATH as described in the CLI guide.
cliptate transcribe interview.mp4 --format srt --output interview.srt
cliptate transcribe recording.wav --format json --download
cliptate tts --text "Hello from Cliptate." --output speech.wav --download
cliptate dictionary add "Cliptate" "MLX"The CLI uses the app's selected models and configuration, launching the app in the background when needed. --download allows missing models to be downloaded. CLI transcription accepts local files and does not create projects or history entries. Run cliptate --help or read the full command reference for batch output, model overrides, and scripting behavior.
- Local speech processing: transcription, VAD, speaker diarization, and TTS inference run on your Mac. Downloaded speech models can be used offline.
- Optional online AI: correction and keyword extraction send text to the provider you configure. Keep these features disabled for an offline speech workflow.
- Translation: Apple Translation uses language resources that may need downloading; availability depends on the language pair.
- Model storage: weights are downloaded separately. Direct builds can reuse the Hugging Face cache at
~/.cache/huggingface/hub; see cache details. - Licensing: the application is MIT licensed. Model weights and dependencies retain their own licenses; see third-party notices.
Requires Apple Silicon, macOS 15+, and Xcode 26+ with Swift 6.2+.
git clone https://github.com/cliptate/cliptate.git
cd cliptate
export DEVELOPER_DIR=/Applications/Xcode.app/Contents/Developer
export CLIPTATE_APP_STORE=NO
swift build
swift test
./scripts/run.shSwiftPM builds the app executable and CLI and runs unit tests. ./scripts/run.sh assembles, signs, and launches the application as Cliptate Dev.app with separate development settings and data. To build the app without launching it, use ./scripts/build-and-run.sh build.
CLIPTATE_APP_STORE=NO selects the free direct-distribution channel, including the CLI and Accessibility-backed dictation. Debug defaults to this channel; Release defaults to App Store restrictions and subscription behavior. App Store billing requires your own local configuration. See the development guide for signing, model caches, and distribution setup.
Bug reports and feature requests belong in GitHub Issues. Include your macOS version, app version, selected model, and reproduction steps; avoid posting private recordings or credentials.
- Contributing — workflow, validation, and commit conventions
- Architecture — code structure and runtime boundaries
- Release notes — version history
- Security policy — reporting vulnerabilities
- Website — product overview