New
Automatic speech-to-text captions
Finish a recording and OneshotFX extracts your mic audio, runs Whisper via Transformers.js right inside the editor, and drops generated captions straight onto your timeline. No cloud, no waiting for uploads — it all happens on-device.
New
Text-to-speech narration from captions
Turn any caption into polished voiceover. Pick a friendly voice — Miles, Ava, and more — and OneshotFX generates narration audio that plays during preview and is baked into your export. Narration can replace your original mic speech while keeping system audio intact.
New
Clear all captions
One click wipes timeline captions, generated narration state, cached STT captions, and related TTS cache — a clean slate without hunting through individual segments.
New
Usage tracking for future freemium limits
OneshotFX now tracks STT transcription count, duration, caption blocks, TTS generation count, character count, and narration duration. TTS cache hits aren't counted as new generations, so you won't burn through limits unnecessarily.
Improved
Better caption timing
Captions now use word-level timestamps and smarter grouping into natural blocks, with improved start/end padding so text appears and disappears at the right moment.
Improved
Improved STT reliability
Longer recordings are transcribed in chunks for accuracy. The engine falls back from WebGPU to WASM automatically when needed, and you can cancel auto-caption generation at any time.
Improved
Narration UX polish
Friendly loading messages during STT/TTS processing, human voice names like Miles and Ava, and a cleaner captions list with per-caption play/refresh buttons removed for narration entries.
Improved
Storage scalability & recording cleanup
Per-recording artifact manifests, centralized artifact deletion, TTS cache ownership by recording ID, and size/LRU-aware cache cleanup. Deleting a recording also cleans related TTS cache entries, and legacy unowned entries are handled safely.
Fixed
Export audio & production build fixes
The Include Audio export toggle now correctly suppresses generated narration as well. STT, TTS, and export workers are properly bundled in production builds so packaged extensions work out of the box.