How to burn captions/subtitles onto a video
Short answer: Ask the AI to caption your video. It transcribes the audio locally with Whisper, writes a subtitle file from the timing, then permanently draws the text onto the picture, all on your machine, nothing uploaded.
Steps
- Get your video into the workspace (drag it into the Files panel, or use the import-file toolbar button). You can just ask in chat (the AI enables media editing itself when needed), but Studio mode has it on by default.
- Ask the AI to caption the clip: “burn captions onto this video.” It transcribes the spoken audio locally into timestamped segments; a long file takes a few minutes.
- Check the transcript the AI shows you (or ask for corrections); the captions are only as good as the transcript. The speech model is English-only, and music, singing, or heavy background noise transcribe poorly.
- The AI writes a subtitle file from the transcript’s timings.
- It burns the captions onto the video, producing a new file with the text baked into the picture: the same kind of hardcoded captions you see on Shorts, Reels, and TikTok. This step re-encodes the video; the audio track is untouched.
- Preview the finished clip in the chat’s media gallery.
Notes & gotchas
- Burned-in captions can’t be toggled off: they’re part of the picture. If you just want the subtitle file itself (to load into an editor, or toggle separately), say so and skip the burn-in step.
- Font coverage: plain Latin text renders reliably; non-Latin scripts (CJK, Arabic) or emoji may show as boxes depending on installed fonts.
- Nothing leaves your machine: transcription and caption burning both run locally through the bundled speech model and FFmpeg.
- Want the clip reframed to vertical too? See how to clip a video into a vertical Short locally: reframe first, then caption, or the other way around, either works.