Voxtral Memos — My First iOS App

vibecoding claude ios ai mistral

Three weeks ago I wrote about publishing my first macOS app. Today I’m following up with my first iOS app: Voxtral Memos.

The idea is simple: record a voice memo, get an instant transcription, and optionally let AI summarize or transform the result. The transcription runs on Mistral’s Voxtral speech model, and the app follows a strict bring-your-own-key model — you plug in your own Mistral API key, the app sends audio directly to their API, and nothing else sits in between. No backend, no middleman.

What It Does

  • Record and transcribe voice memos with support for 13+ languages
  • Smart summaries — extract bullet points, todo lists, or journal entries from your recordings
  • Custom templates — write your own prompt templates for any transformation you need
  • Multilingual — both the app and its built-in templates are available in English, German, French, Spanish, Italian, Portuguese, and Russian

What It Doesn’t Do

  • No account required — no sign-up, no login, no user profiles
  • No tracking — zero analytics, no telemetry, no third-party SDKs
  • No cloud storage — everything stays on your device in SwiftData
  • No subscription — the app is free, you only pay for what you use via the Mistral API (typically less than a dollar per month for casual use)

The API key is stored in the iOS Keychain. Audio only leaves your device when you explicitly request a transcription. There is no backend server.

Why Mistral

I wanted a European AI provider with a strong speech model, and Mistral fits. They’re GDPR-compliant, transparent about data handling, and don’t use your inputs for training. Voxtral handles multilingual audio well, and the API pricing is reasonable — about $0.01 per five-minute transcription.

Try It

Voxtral Memos is free and open-source. The app is on the App Store, and the source code is on GitHub.