AetherCut^™ auto-captions

AetherCut's auto-caption tool generates word-level timestamps using OpenAI Whisper. You can edit each word inline, restyle the captions (Submagic-style animated word-by-word reveal is available on Pro), and either burn them into the export or download a sidecar .srt / .vtt file.

Honest disclosure that 90% of caption tools never make: this is one of the few AetherCut features that genuinely uses a server-side API. Your audio (not your video — just the audio track) is sent to OpenAI's Whisper endpoint for transcription. Whisper returns text + timestamps. Nothing else leaves your device.

If you'd rather your audio never leaves either, AetherCut's Privacy Mode toggle disables the feature entirely. The rest of the editor — including the on-device timeline, background remover, color grading, video stabilization, green-screen — continues to work normally.

What does AetherCut send for auto captions?

AetherCut sends only the clip's audio track to Whisper in a single request. It does not send video frames, the project file, or other clips. The returned transcript includes word level timestamps for frame accurate caption placement.

Only the audio track of the clip you're captioning is sent to Whisper. Not the video frames, not your project file, not your other clips. The audio is sent as a single request, transcribed in one round-trip, and OpenAI's published Whisper API policy does not retain the audio after the response is returned.

The transcription includes per-word timestamps. We use those timestamps to position captions on the timeline frame-accurate. The word-level granularity is also what enables features like AI Silence Removal and Voice Cleanup (disfluency cuts) — both of which then run locally on the transcript.

Which languages does AetherCut support for auto captions?

AetherCut surfaces 20 languages measured to work well in production, including English, Spanish, French, German, Japanese, Mandarin, Arabic, Korean, Hindi, and more. It can also auto translate the transcript into another language separately.

Whisper supports about 99 languages with widely varying accuracy. AetherCut's UI surfaces the 20 we've measured working well in production: English, Spanish, French, German, Japanese, Mandarin, Portuguese, Arabic, Korean, Hindi, Italian, Russian, Turkish, Dutch, Polish, Greek, Vietnamese, Indonesian, plus Hindi-Latn transliteration.

You can also auto-translate the resulting transcript into a different language as a separate step. For voiced dubbing rather than just translated text, AetherCut's AI Dubbing tool (Pro, ElevenLabs) ports the spoken audio into 25 target languages.

Can AetherCut export captions as SRT or burned in video?

AetherCut can export captions burned into the MP4 or as separate SRT and VTT files. Free includes static block captions, while Pro adds animated captions with local Canvas based rendering after transcription.

The free tier ships static block captions — the standard YouTube-style two-line bottom-third rendering, restylable for font, color, outline, and background.

On Pro, animated captions (one-word-at-a-time reveal with emphasis on certain words, the format popularized by Submagic) are available. The animation runs on Canvas locally — only the original transcription was the API call.

Both export burned-in (rendered into the MP4) or as a sidecar .srt / .vtt file you can ship separately to YouTube, social platforms, or accessibility tooling.