Follow playback through the script, with clearer issue cards and a faster History list.
As audio plays, the current sentence is highlighted in the script and scrolls smoothly toward the center. Short scripts stay still when they already fit on screen; after you scroll or select another passage, choose Resume auto-scroll to follow playback again.
Extra-speech cards keep English words whole and use nearby punctuation to make Chinese excerpts easier to read.
A built-in sample project, Settings in its own window, and automatic re-review of extra speech.
Try a sample: on the empty import screen, open a short ready-made project with a first take containing four planted mistakes and a corrected second take. It does not start the free trial and is never saved to History.
When extra words found in one take are gone in the next, the issue is now marked as no longer detected instead of always asking you to listen again.
Settings, History, Feedback and Activation open in a regular window with the standard close button; ⌘W and Esc close it.
Pauses in the waveform show as a light line instead of gaps, the waveform is centred, and the playhead is easier to see.
Recognition models and settings are unchanged. Updating from 1.1.0 or 1.1.1 downloads a small patch instead of the full app.
Listen plays the right spot in VBR MP3 files, and long silence no longer turns into invented text.
In variable-bitrate MP3 files, Listen could start away from the marked spot, by up to several seconds in long recordings. It now starts exactly where the issue is. This fixes the known limitation listed in 1.1.0.
The speech recognizer could produce phrases such as “Thank you” in long silence, usually at the end of a file, and flag them as extra or misread speech. Text found only inside two seconds or more of near-silence is now ignored, and a fully silent file reports that no speech was recognized.
The Save markers button in the Export panel now matches the other buttons.
Recognition models and settings are unchanged. Updating from 1.1.0 downloads a small patch instead of the full app.
Save issues as DAW markers, and check longer recordings.
Save markers: export confirmed issues as markers for REAPER, Adobe Audition or Pro Tools (via a MIDI file). Issues with a time range become regions, and numbering matches the fix list. REAPER import has been tested; Audition and Pro Tools follow their documented import formats.
Scripts and audio of about 1–3 hours can now be compared, and long reviews save and reopen much faster in History.
Fixes a crash that could quit Proof Take during a check.
A check no longer quits the app partway through. The bundled speech engine (whisper.cpp) could stop when a short stretch of audio at the end of a recognition window produced text; 1.0.1 includes the upstream fix.
Recognition settings, models and your review history are unchanged.
Fixes app updates that could stay at “Installing…”.
Installing an update from Settings no longer gets stuck at “Installing…”. Proof Take saves your review, closes Settings, then restarts to finish the update.
If another dialog is open, the downloaded update is kept so you can try again.