Free, privacy-first speech-to-text — local on desktop, secure cloud on mobile.
| Workspace | Transcript & exports |
|---|---|
 |
 |
- 100% free — no sign-up, no paywall
- Desktop-local transcription — audio stays in the browser; English uses Moonshine Base and other languages use Whisper Small
- WebGPU first, WASM fallback — local inference works across modern desktop browsers
- Mobile transcription — audio is processed temporarily in the cloud and removed after the result is returned
- Long recording support — chunked processing, retry/backoff, cancellation, progress, and ETA
- 12 languages / 9 formats — MP3, WAV, M4A, MP4, OGG, FLAC, AAC, WEBM, and OPUS
- Cross-platform flag selector — local SVG flags render consistently on Windows, macOS, iOS, and Android
- Export — plain text, timestamped text, JSON, Word (.docx), and Unicode-aware PDF
- Continue with AI: one-click open transcript in ChatGPT, Claude, Gemini, or Grok
- English and Turkish UI — localized application, metadata, structured data, and privacy pages
- Privacy-first lifecycle — temporary mobile audio is cleaned up automatically
English, Turkish, Spanish, French, German, Italian, Portuguese, Russian, Arabic, Hindi, Japanese, and Korean.
1. Select language → 2. Upload audio → 3. Review and export
12 languages 9 file formats Text, JSON, Word, PDF
Audio is decoded and resampled in-browser → a Web Worker loads the AI model → overlapping windows are transcribed sequentially → segments are merged and deduplicated. The audio file is never uploaded.
- English → Moonshine Base (fast, lightweight model)
- Other languages → Whisper Small (WebGPU with WASM fallback)
- Local recordings are limited to 2 hours and guarded by a safe browser decode-size limit
Audio is prepared in the browser, split into mobile-safe chunks when needed, and sent for temporary cloud transcription. Results are merged with timestamps and the temporary audio is removed automatically. Progress, retries, cancellation, and recordings up to 2 hours are supported.
flowchart LR
A[Select language and audio] --> B{Desktop or mobile}
B -->|Desktop| C[Local browser AI]
B -->|Mobile| D[Temporary cloud transcription]
C --> E[Timestamped transcript]
D --> E
E --> F[Copy, JSON, Word, PDF, AI chat]
| Layer | Technology |
|---|---|
| Framework | Next.js 16 (App Router) |
| UI | React 19, Tailwind CSS 4, Radix Select, Lucide icons |
| Language | TypeScript (strict) |
| AI Models | Whisper Small, Moonshine Base, Whisper Large V3 |
| Document Export | docx, FileSaver, jsPDF |
| Testing | Vitest and ESLint |
pnpm install
pnpm devOpen http://localhost:3000. Desktop transcription works without any API keys.
Useful checks:
pnpm test
pnpm lint
pnpm build- Desktop: 100% local — audio never leaves your browser
- Mobile: temporary cloud processing with automatic upload cleanup
- No accounts or analytics trackers