Whisper Transcription

■ New • Import from URL: paste a link to an audio or video file, and transcribe it straight from the home screen • Added Claude Opus 5 to the Anthropic model picker for summaries, titles, and transcript chat ■ Improvements • Redesigned Settings with new icons • Assistant transcription now runs on Mistral's Voxtral model; larger files are still handled by ElevenLabs Scribe • Clearer errors when an on-device model download fails because the connection is blocked or unstable • Updated German, Dutch, and French translations ■ Bugfixes • Fixed the detected language not showing on transcripts made with the Assistant • Fixed some audio files being rejected by the Assistant because the upload did not include the original filename • Fixed the "file too large" message from the Assistant not being translated

■ New • Added export previews, with support for VTT and Markdown subtitles and DOTE files • Added Anthropic API key support and model selection for summaries, titles, and transcript chat • Added an OpenAI model picker, including support for custom models • Transcripts can now be renamed manually ■ Improvements • Redesigned transcript chat with clearer loading and empty states, a better input bar, and easier response copying • Improved the transcript detail layout on iPad, including a wider playback bar and a collapsible inspector • Improved search result reliability and transcription retry actions • Improved the confirmation shown when deleting all dictations • Updated translations throughout the iPhone and iPad apps ■ Bugfixes • Fixed the selected AI service sometimes changing when it was temporarily unavailable • Fixed opening Custom API Keys unexpectedly switching the selected AI service to OpenAI • Fixed unclear errors when an on-device model download became corrupted • Fixed sharing a transcript chat response sometimes sharing the prompt instead of the generated answer

■ Improvements • A cleaner, more consistent look across the app, including a redesigned on-device model download screen. • Transcribe audio-only MP4 files. • Smoother search on the home screen. • Sharing a video now shows "Share Video" with a video icon, audio files show audio. • More room on iPad: the home screen shows more of your history with longer, multi-line previews that fill the screen. • Quick paste buttons on the OpenAI and ElevenLabs API key fields. • VoiceOver and accessibility improvements throughout - progress indicators and the playback scrubber now announce what's happening (for example "1:23 of 4:56"). ■ Bugfixes • Fixed short audio clips sometimes producing an empty transcription. • Fixed the Share Sheet prompt on the home screen dismissing when you didn't mean it to.

■ New • Local transcription models now run on latest iPad models ■ Improved • Assistant can now handle files twice as long

■ Improvements • Improved stability when using local models