3.2.7 - Fix crash for devices without Metal support & revert to CPU - Upgrade to Whisper.cpp 1.5.1 3.2.6 - Add support for Cantonese for Large V3. - Fix Hebrew language bug. 3.2.4 - Support for Whisper 1.5 with Metal and removal of CoreML. Big performance improvement over CoreML but a massive speed improvement for Medium and Large (which didn't have CoreML). On an M1 Max a 400% increase for Large. On an iPhone 14 Pro 100% increase for Medium. - Support new Large-V3 model from OpenAI. - Support beam search size of 0 (= Greedy). - New fast abort mechanism (will abort or stop immediately where before it could take a long time when using big model).

3.2.6 Add support for Cantonese for Large V3. Fix Hebrew language bug. 3.2.4 - Support for Whisper 1.5 with Metal and removal of CoreML. Big performance improvement over CoreML but a massive speed improvement for Medium and Large (which didn't have CoreML). On an M1 Max a 400% increase for Large. On an iPhone 14 Pro 100% increase for Medium. - Support new Large-V3 model from OpenAI. - Support beam search size of 0 (= Greedy). - New fast abort mechanism (will abort or stop immediately where before it could take a long time when using big model).

3.2.5 - Support for Whisper 1.5 with Metal and removal of CoreML. Big performance improvement over CoreML but a massive speed improvement for Medium and Large (which didn't have CoreML). On an M1 Max a 400% increase for Large. On an iPhone 14 Pro 100% increase for Medium. - Support new Large-V3 model from OpenAI. - Support beam search size of 0 (= Greedy). - New fast abort mechanism (will abort or stop immediately where before it could take a long time when using big model).

3.1.7 Transcribe video from the Photos app on iOS and macOS. Fix memory leak when job finishes and doesn't release transcriber.

3.0.6 Bottom toolbar icons are better spaced apart. Pro downloads will be disabled when Pro not purchased. 3.0.5 Model configs: A model config (configuration) is a Whisper model combined with transcription and rendering options. Previously, you could only set the Whisper model, CoreML enabled etc. in Settings. This meant that changing the settings was awkward, and you could not change them while doing a transcription. Now, you can create multiple configs and use them for different tasks and change them while transcribing. Changing the config will restart the transcription. This means now you can: • Have a "fast" default config to quickly transcribe something and change to a "accurate" config when you want a more accurate result. • Tweak a config during transcription (e.g. turn translation on/off or change the audio language). • Set your favourite config as the default and copy an existing config to tweak it. Two new advanced options have been added: • Per-word timestamps. *This is experimental*. Each word will have a timestamp. • Beam search size. A smaller beam search is faster but will create less accurate results. Before 3.0 Hello Transcribe used the "greedy" search method, which is the same as a search size of 1. Other: • App will not go to sleep while a transcription is in progress.

Model configs: A model config (configuration) is a Whisper model combined with transcription and rendering options. Previously, you could only set the Whisper model, CoreML enabled etc. in Settings. This meant that changing the settings was awkward, and you could not change them while doing a transcription. Now, you can create multiple configs and use them for different tasks and change them while transcribing. Changing the config will restart the transcription. This means now you can: • Have a "fast" default config to quickly transcribe something and change to a "accurate" config when you want a more accurate result. • Tweak a config during transcription (e.g. turn translation on/off or change the audio language). • Set your favourite config as the default and copy an existing config to tweak it. Two new advanced options have been added: • Per-word timestamps. *This is experimental*. Each word will have a timestamp. • Beam search size. A smaller beam search is faster but will create less accurate results. Before 3.0 Hello Transcribe used the "greedy" search method, which is the same as a search size of 1. Other: • App will not go to sleep while a transcription is in progress.

Transcribe MP4 video files.

Fix macOS microphone issue. Add missing 3rd party library licenses.

Multiplatform support (macOS and iPadOS) using the same in-app purchase for Pro. Better support for different audio formats.

Fix broken toolbars when users have Display Zoom enabled

Fixed a bug where the result would always show that the "tiny" model was used even when using other models.

VTT and SRT file export for captioning video. Your device will not auto-lock while transcribing. Improved progress for file transcription. Improved Dictation stability. Fix a Scroll Lock cutting off the bottom timestamp. Fix Restore Purchase workflow.

Bugfix for the dictation "bugfix" :/ Sorry about that

Interim improvement for memory leak, long-running transcriptions should be more stable now. Proper fix coming asap. If you have issues you can disable CoreML.

- Timestamps - New toolbars - On-demand model loading - Rewritten Whisper framework - Various small improvements & bugfixes

- CoreML can be disabled for iPhones that don't have enough memory to optimise the largest CoreML models. - Show minimum recommended iPhone during Pro Purchase. - Added warning about loading larger CoreML models which may crash the app. - Loading screen will only show after a small delay to avoid flashing.

Huge performance improvements using CoreML - users will see 3x - 7x faster transcriptions.

• Shortcuts for dictate and transcribe from file. • Once-off Pro purchase as an alternative to subscriptions. • Pause, resume and clear dictation. • Scroll lock - the result will automatically scroll to the latest segment. • Document picker - you can now select audio document directly from the app. • Remove non-speech phrases like [Applause] from the result (Pro only). • Change the transcription font size, which is useful when using the app for live dictation (Pro only). • Less cluttered and simplified UI. • Raw audio is not saved (this lead to the app crashing when saving large transcriptions).

2.0.5:  - Bugfix when installing app for first time and Info view breaks Dictation. 2.0.4: - 5 more Whisper models to choose from. - Saving transcriptions on your device and in iCloud. - Multi-lingual support with bigger models and switchable translation to English. - In-app purchase of a subscription or transcription minutes for larger models.

5 more Whisper models to choose from. Saving transcriptions on your device and in iCloud. Multi-lingual support with bigger models and switchable translation to English. In-app purchase of a subscription or transcription minutes for larger models.

Bugfix for Signal Voicenotes

Dictation! You can now dictate and it will be transcribed for you in real time.

Support for Telegram voice notes (Ogg Opus). Support for Ogg Vorbis audio. Support for Ogg Opus audio. Error messaging when things go wrong unexpectedly. A new cancellation dialog when waiting for background transcription process to finish.