Maina Voice // AI Voice Dictation & Engine Benchmarking
Maina Voice is a local-first, browser-based speech-to-text workbench. Record audio in your
browser, compare OpenAI GPT-Transcribe, Deepgram Nova-3, NVIDIA Parakeet, Fish Audio and Groq
Whisper, and keep every recording, transcript and setting in your own IndexedDB. Your audio is
sent directly from your device to the provider you choose using your own API key. There is no
account, no analytics beacon and no first-party backend that stores your data.
What you can do without signing up
- Record from your microphone or upload an existing audio file.
- Run the same clip through multiple speech-to-text engines and compare speed, cost and quality.
- Track every transcript edit and re-transcription as a versioned history.
- Export your entire library as a ZIP backup, or import it again on any device.
Compare engines side by side
The Compare view benchmarks OpenAI GPT-Transcribe, Deepgram Nova-3, NVIDIA Parakeet TDT v3,
Fish Audio Transcribe-1 and Groq Whisper Large v3 Turbo on the same recording so you can see
latency, estimated price and output quality before choosing a provider for your workflow.
Privacy by design
Recordings, WAV blobs, generated transcripts, translations, comparison results and application
settings live in the IndexedDB inside your own browser. Maina Voice never proxies, logs or
stores your audio on its own servers.
Get started
Enable JavaScript for the full app, or read the About, Privacy and Contact pages to learn more
about the project and how your data is handled.
