Word-Level Generated Captions
Generate caption clips from precise word timestamps so spoken words appear at the right moment on the timeline.
ScreenKite makes captions from your speech. The transcription engine gives a time for each word. ScreenKite groups the words into caption lines, and it can highlight each word as it is spoken.
Choose a Transcription Engine
Open Settings → Captions. The page has three tabs: AI, AI Text Model and Formatting & Export.
In the AI tab, open Advanced → Transcription engine and choose an engine:
| Engine | Where it runs | Notes |
|---|---|---|
| Qwen3 (on-device) | On your Mac | The default free local engine. Downloads its model once, then works offline with precise word times. |
| WhisperKit | On your Mac | Download a model with Download Selected Model. |
| MOSS Transcribe + Speakers | On your Mac | Adds speaker labels. |
| Apple Speech (on-device) | On your Mac | Needs macOS 26 or later. Choose the spoken language. |
| ElevenLabs | Online | Best for precise word timings. Needs an API key. |
| Fish Audio | Online | Needs a funded API account. |
The AI Text Model tab sets the model for AI text work such as cleanup and proofreading: OpenAI, Azure OpenAI, Anthropic or Google Gemini. It does not set caption timing.
Test an ElevenLabs key
After you enter a key, click Test Key. The result shows next to the button:
| Result | Meaning |
|---|---|
| Valid for Speech to Text | The key works. |
| Invalid API key | The key is wrong or revoked. |
| Key needs ElevenLabs Speech to Text permission | Turn on Speech to Text access for the key in the ElevenLabs dashboard. |
| ElevenLabs rate limit reached. Try again later. | Wait, then test again. |
Generate Captions
- Open the project in the editor.
- Click the Captions tab in the inspector.
- Click Transcribe.
If transcription is not set up, the button shows Set up transcription. ScreenKite opens the Transcription setup window to help you choose an engine.
The captions go on the Captions track of the timeline. If ScreenKite finds no speech, it shows No speech detected.
Edit and Style Captions
The Captions tab has three tabs:
- Transcript: read the transcript and fix spelling. Spelling fixes do not change the video.
- Style: choose a caption style, such as Pill, Plain, Karaoke or Boxed. Set Text, Box, Position and Lines. Turn on Highlight spoken word to highlight each word as it is spoken.
- Edit: cut words, filler words and silences. See Text-Based Editing.
Line length
In Style → Lines, set Breaks and Max characters, then click Apply Line Length. The default is 42 characters per line, and 18 for Chinese, Japanese and Korean.
Export Captions
- Choose File → Export Captions....
- Or in the export dialog, choose Subtitles only and pick SubRip (.srt) or WebVTT (.vtt).
To copy chapters for YouTube, open the Transcript tab menu and choose Copy YouTube chapters. Or in the Markers tab, click Copy Chapters.
Agents and Captions
An AI agent can make captions with the same engine. Ask it in the AI Chat Assistant panel, or from a terminal. See Agentic Video Editing.