# Captions & transcripts: Captions that write themselves, on your Mac.

> On-device transcription, word-by-word captions and transcript-based cutting.

Canonical: https://www.smoothcapture.app/features/captions

Transcribe a recording, a microphone track or a voice-over without uploading it. Captions highlight each word as it is spoken, and the transcript doubles as a way to cut the video.

## Highlights

### Runs on your Mac

Audio is transcribed locally. Nothing is sent to a server.

### Two engines

Apple Speech on macOS 26, or FluidAudio on any supported macOS after a one-time model download of about 200 MB.

### Word-level highlighting

Each word lights up as it is spoken, karaoke style, with adjustable caption size.

### Pick the source

Transcribe the screen recording’s audio, your microphone track or an audio clip such as an AI voice-over.

### 25 languages with FluidAudio

English plus 24 European languages, including German, French, Spanish, Italian, Portuguese, Dutch, Polish and the Nordic languages.

### Cut by deleting text

Delete a sentence in the transcript editor and the matching video is removed. A filter finds filler words.

## How to do it

### 1. Choose the source and language

Open Subtitles in the inspector, pick the source and language, then click Transcribe Audio.

### 2. Show and style the captions

Turn on Show Subtitles and set the size. Words are highlighted in time with the speech.

### 3. Fix and cut from the transcript

Open the transcript editor to correct words, filter for fillers and delete sentences. Re-generate the transcript at any time.

## From the CLI or an AI agent

An agent can transcribe a clip or a narration track and read back word timings mapped to the timeline. These words came from the AI voice-over in the demo above.

```bash
smoothcapture cli editor transcribe --input - <<'JSON'
{"audioClip": "F5CF9040-…", "locale": "en-US",
 "revision": "…"}
JSON
# → words: When you talk, the music steps aside.
```

## FAQ

### Does transcription upload my audio?

No. Both engines run on your Mac. FluidAudio downloads its models once; after that it works offline.

### Which macOS do I need for captions?

Apple Speech needs macOS 26. On macOS 15.2 and later you can use FluidAudio after a one-time download of about 200 MB.

### Which languages are supported?

FluidAudio supports English and 24 European languages, from German, French and Spanish to Polish, Ukrainian, Czech and the Nordic and Baltic languages. Pick the language in the Subtitles tab before transcribing.

### Can I edit the captions?

Yes. Correct words in the transcript editor, or re-generate the transcript after changing the source or language.

### Are captions included when an agent exports from the CLI?

Not yet. Headless CLI and MCP exports skip the subtitle track. Export from the app to burn captions in, or use editable titles for captions in scripted exports.

## Related

- [Timeline editor](https://www.smoothcapture.app/features/timeline-editor)
- [Audio, music & voice-over](https://www.smoothcapture.app/features/audio-and-voiceover)
- [Titles & motion graphics](https://www.smoothcapture.app/features/titles-and-graphics)