CaptionCut

Turn spoken words into local captions and reversible edits. CaptionCut includes its multilingual speech model and tokenizer so the bundled recognizer works without a model download or developer-operated processing service.

Preparing for the App Store. Not yet available for download.

Tools for your own media.

Included local speech model

Multilingual recognition and tokenizer files are bundled. No remote model or tokenizer fallback.

Words linked to edits

Search and correct timed words; select their media ranges, undo edits or restore cuts.

Captions for delivery

Retimed SRT/VTT, edited AAC audio and H.264 captioned videos in shareable ZIP packages.

Saved work

Draft text, timings, cut history and completed exports recover after reopening.

What is included

An optional Apple offline recognizer is available only when its language works locally on the device and speech permission is granted. The bundled recognizer does not need that permission. Accuracy varies with language, accents, noise and overlapping speech. Review text, timings and cuts before sharing. The bundled model is compact; it does not guarantee verbatim or professional transcription. Video export uses H.264 MP4 and foreground processing.

One purchase includes all tools. No account, ads, subscription or in-app purchases. Imported recordings are not uploaded by the app. Cloud-only imports and a share destination you choose may require internet.

Processing and exports work locally after import. Cloud-only Photos or Files, support email and online sharing destinations may need internet.

Prepared update: Create local captions, correct their text and timing, and prepare edited media with reviewable subtitle alignment

Prepared update for version 1.0 build 3. Availability remains subject to App Store review and release. These notes describe the prepared workflow; check the installed version before following them.

Timing desk maps two existing cue times to two spoken moments you identify in the recording. Listen through the waveform and player, capture playback positions, inspect before-and-aligned cues and audition their corrected starts. Apply the alignment through the real edit history, then export aligned SRT, VTT and edited audio. Saved anchors, applied alignment history and captions reopen with the project.

Editor and Subtitle QA retain on-device transcription, text correction, word-selection cuts, undo and delivery tools. The included Whisper model and tokenizer run locally; optional Apple recognition requires a locally supported language and permission.

Alignment follows your listening and selected anchors. It is not automatic proof of synchronization. Invalid, overlapping or clipped cue mappings are blocked. Transcription accuracy varies; review the text and media before sharing.