iOS
Every Scribe0 for iOS release, in plain language.
Download on the App Storev1.0
Scribe0 for iPhone is out on the App Store. On-device dictation, free, with no account.
Added- Email a note straight from the record screen. The new envelope button opens a pre-addressed email with the note properly formatted (headings, bold text, lists), fixing the wall-of-text emails that some mail apps made of shared notes. Save your usual address once in Settings under "Email note to" and it fills in automatically. The email is written and sent from your own mail account, so the note never passes through Scribe0 servers. The regular share button works exactly as before.
Changed- The app is now simpler and completely free, always. Cloud enhancement runs Claude Haiku with 20 enhances a day included, the model picker is gone, and the optional bring-your-own-key OpenRouter backend has been removed. Scribe0 Cloud (Australian AWS servers) is the only place transcript text can ever be sent, and only after you approve it. On-device enhancement stays unlimited.
- The included prompts now use Scribe0 Cloud out of the box, so results are strong from the first enhance. The first cloud enhance asks your permission before any transcript leaves the device; switch any prompt to On-device to keep everything local.
- New Settings screen "Data and AI" spells out exactly what the AI features send, where it goes, and what never leaves your device. Onboarding now names the AI service up front.
v1.0 build 13
Changed- Settings is reorganised around a single LLM choice. A three-way switch (Scribe0 Cloud, On-device, OpenRouter) sets the default for new prompts, with only the chosen backend's options shown below it. The same switch appears in the prompt editor, so per-prompt overrides look and work the same way. The internal endpoint field is gone and the descriptions are shorter.
- Speaker labels are now off by default. Turn them on with the toggle above the prompt picker when you record a conversation.
- The app now says "speech model" in everyday screens instead of the model's technical name. The full model name (Parakeet TDT-CTC 110M) still appears under Device capabilities and in diagnostics.
Fixed- The speech model now downloads reliably on first launch. It is served from our own content network as a single verified file, instead of a third-party model host whose transfer could stall and leave the setup screen loading forever on some networks and devices. The first-run screen now shows a live download percentage, and a failed or interrupted download offers a Retry button instead of spinning indefinitely. The whole model, including a component the app previously tried to fetch separately on every launch, is now bundled into that one download, so preparing on-device transcription makes no other network calls.
- The model is downloaded only once. Later launches load the already-downloaded model from the device, and the setup screen now says "Loading" rather than "Downloading" so the one-time download and the every-launch load into memory read as two clearly separate steps. The downloaded model is also kept out of device and iCloud backups, since it can always be fetched again.
v1.0 build 9
Changed- Scribe0 is now iPhone only. The iPad layout was never tested, so shipping it would have meant asking you to use a screen we had not designed for.
v1.0 build 8
Added- AI Enhance works with no setup. The cloud enhance step is free out of the box, with no account and no key to paste. A Pro sign-in unlocks the larger models and removes the daily limit.
- Prompt gallery. Browse and import community prompts.
- Prompts can now be deleted.
v1.0 build 6
Security- Your recordings and transcripts are encrypted at rest, with an optional Face ID or Touch ID lock on the app itself.
Fixed- An earlier build shipped without its security entitlements, so data protection and the shared container were not actually active on device. The release build now signs correctly and the entitlements are verified in the shipped app. If you were on a build before this one, update.
v1.0 build 4
Added- The screen now stays awake while you are recording.
- A clear loading overlay while the speech model loads, instead of an app that looks frozen.
Fixed- Audio file import now decodes reliably.
v1.0 build 1
First TestFlight build. On-device dictation for iPhone.
Added- Local speech to text on the Neural Engine (Parakeet). Audio never leaves your iPhone.
- AI Enhance for turning dictation into a structured note.
- Import an audio file and transcribe it.
- Prompt presets and session history.