AI Transcribe and AI Captions

Two Smart AI wizards turn speech in any video or audio layer into an editable transcript — and, with AI Captions, straight into styled, voice-synced caption layers.

Run them

Select the layer and open Smart AI → AI Transcribe… or AI Captions… in the timeline toolbar. Both are three-step wizards:

1. Setup — pick the transcription engine (Automatic — OpenAI first with WaveSpeed fallback, OpenAI Whisper, or WaveSpeed Whisper), the language, and whether to keep filler words. Long files are handled: the server extracts a compact speech track first, and speech beyond the transcription ceiling is transcribed in slices and stitched.
2. Progress — a progress rail with phases and a running log (the bar is labelled estimated — transcription is a single request).
3. Results — review the transcript before anything touches the timeline. Choose whether it lands on an extracted Voice take (the speech is lifted onto its own Voice channel, editable word-by-word in the Transcribe Editor) or just attaches to the layer as-is.

A layer that already has a transcript offers Use saved, so restyling captions never pays for the API call twice. Saved transcripts appear in the layer panel’s AI Data list — with word count, language and caption count — and can be removed from there.

AI Captions — the full animation catalog

The AI Captions results step is a 50/50 split: caption preview, timing strip and transcript on the left; the complete Text Animations picker on the right — all caption presets with live previews, search and category filters. The preview runs the same renderer the preset rows use, fed this caption’s real words.

• Split & Timing — split by phrases or a fixed word count; position bottom / center / top.
• Sync to voice (on by default) — each word appears and highlights exactly when it’s spoken, from the transcript’s per-word timestamps.
• Exit — untick to keep the entrance and let the text simply cut out.
• Text size and legibility shadows — captions open with Enable text shadows already ticked, because white text over bright footage disappears without one.

Applied captions are ordinary text layers — see Captions for editing them.

Where to go next

• Captions
• Voice Channels and Transcribe Editor
• AI Shorts