Maina Voice

Maina Voice

Local-first, browser-based speech-to-text workbench. Dictate, transcribe and benchmark cloud voice models without handing your recordings to a third-party server. All recordings, audio blobs, transcript versions and settings live in your browser's IndexedDB. The only network call is the one you choose to make, from your browser directly to OpenRouter or Groq with your own API key.

Dictate & transcribe

Record from your microphone or upload an audio file. Audio is normalized to WAV and sent straight to the provider you pick.

Benchmark engines

Run the same clip through OpenAI GPT-Transcribe, Deepgram Nova-3, NVIDIA Parakeet and Fish Audio. Compare latency and accuracy side by side on your own voice.

Transcript history

Every edit is versioned, so you can track how a transcript changed over time.