Skip to content
Drop 001Hoodie is liveShop the hoodieShop
BridgeMind
Download
BridgeVoiceIncluded with Pro

Your next thought.
Already text.

Speak a coding prompt, a message, a moment of clarity. BridgeVoice puts your words where you’re already working.

Download BridgeVoiceTry the app demo
  • Local or cloud engine
  • Up to 100 languages
  • Mac, Windows, Linux

Try BridgeVoice in your browser

BridgeVoice

Home

What you've shipped by voice.

Ready to dictate

Try a sample here. In the desktop app, hold your key and speak.

Fn / Globe
Local · Parakeet V3
Words
44
Speaking time
21s
Recordings
2
Avg pace
126 wpm

Recent activity

The new onboarding flow is ready to review. I have checked the keyboard navigation and mobile layout. Let me know what you think.

10:24 AM · 23 words · 11s

An idea for tomorrow: let builders search the model list by language, so choosing the right one takes a single step.

9:41 AM · 21 words · 10s

Simulated dictation with sample text. No microphone access, uploads, or model downloads.

  1. 1Hold your shortcut.

    Stay in the text field you’re already using.

  2. 2Say what’s on your mind.

    A brief, a reply, or the next big idea.

  3. 3Release. Keep going.

    Your speech lands as editable text.

Your voice.
Your choice of engine.

Keep speech recognition on your computer, or let a cloud model handle it. The choice is yours in BridgeVoice Settings.

Local transcription

Your audio. Your computer.

On device
  1. Your microphoneRecorded on this machine
  2. Your computerModel runs on your hardware

Download a model once. Speech recognition then runs on your hardware, with recorded audio staying on the device.

Available modelsDownload
Parakeet V3NVIDIA · 25 languages
671 MB

A good place to start for everyday dictation. Automatically detects supported languages. Available on Apple silicon, Windows, and Linux. On Intel Mac, choose Whisper Turbo Q5.

Read the model reference
Whisper Turbo Q5OpenAI · quantized · Multilingual
574 MB

Broad language coverage in the smallest download in the current catalog. The default local choice on Intel Mac. A quantized version of Whisper Large v3 Turbo.

Read the model reference
Distil-Whisper v3.5Distil-Whisper · English only
1.52 GB

An English-focused alternative for local speech recognition. Choose this for English dictation. It is not a multilingual model.

Read the model reference
Whisper Large v3OpenAI · Multilingual
3.10 GB

The full-size Whisper model for broader local language needs. The largest download in the catalog; allow more memory and processing time.

Read the model reference

Download sizes are approximate, not memory requirements. Performance depends on your hardware and recording.

Cloud transcription

Less work for your computer.

Hosted
  1. Your microphoneRecorded on this machine
  2. BridgeMind APIRelays the audio
  3. Hosted providerTranscribes and returns text
Microsoft AI

MAI-Transcribe-2

Hosted through OpenRouter

  • 60 supported languages
  • Automatic language detection
  • Mixed-language speech recognition
  • No local model download

Audio is sent through the BridgeMind API to the hosted provider. An internet connection is required.

Explore the cloud model

Transcription and rewriting are separate choices. Optional Polish and Enhance Prompt send transcript text to hosted AI and use BridgeMind credits, even with a local transcription model. Account access and connected features may also need the internet.

Privacy details

Does it speak
your language?

Find your language, then choose your model. Language coverage varies between local and cloud transcription.

Model references checked September 9, 2026

60 languages · MAI-Transcribe-2

  • Afrikaansaf
  • Arabicar
  • Armenianhy
  • Assameseas
  • Azerbaijaniaz
  • Bengalibn
  • Bosnianbs
  • Bulgarianbg
  • Cantoneseyue
  • Catalanca
  • Chinese (Simplified)zh
  • Czechcs
  • Danishda
  • Dutchnl
  • Englishen
  • Estonianet
  • Filipinofil
  • Finnishfi
  • Frenchfr
  • Galiciangl

60 supported languages, detected automatically. The desktop’s manual picker shows a subset; choose Auto-detect for the full model coverage.

Model language reference

Language support describes speech recognition, not app interface translation. Accents, background noise, and specialist terms can affect results. Review your transcript before sending.

Stay with the idea.
Let your voice do the typing.

Give your coding agent the context it needs. Write a thoughtful reply. Capture a note before the thought slips away.

A guide to better spoken coding prompts
Spoken in 10 secondsCoding prompt
“Add an empty state to the dashboard. Keep the current design, make the next step clear, and check it on mobile.”
Ready to review. Ready for your agent.

A few things
worth knowing.

Models, privacy, languages, and the details before your first dictation.

Ask the builder community

What is BridgeVoice?

BridgeVoice is a desktop dictation app for macOS, Windows, and Linux. Hold your configured shortcut, speak, then release to transcribe and place text in the field you are using. It works for coding prompts, messages, notes, and everyday writing.

Can I use BridgeVoice offline?

Local speech recognition runs on your computer after you download a model. Account access, model downloads, cloud transcription, and connected features can still require the internet. Local dictation does not mean every part of the app works offline.

Which local model should I choose?

Start with Parakeet V3 if your language is among its 25 supported languages and your computer supports it. On Intel Mac, start with Whisper Turbo Q5. Whisper Turbo Q5 and Large v3 offer broader multilingual coverage; Distil-Whisper v3.5 is English only. Speed and accuracy depend on your computer, microphone, language, and recording.

Which model does cloud transcription use?

BridgeVoice cloud transcription uses Microsoft MAI-Transcribe-2 through OpenRouter, via the BridgeMind API. It supports 60 languages and automatic language detection. Cloud mode needs an internet connection and sends your audio for hosted processing.

How many languages does BridgeVoice support?

Language coverage depends on the selected model. MAI-Transcribe-2 supports 60 languages, Parakeet V3 supports 25, and Distil-Whisper v3.5 supports English. Whisper Large v3 and Turbo have a 100-language recognition vocabulary, with accuracy varying by language. Use the language explorer on this page for the complete lists.

What if my language is missing from the cloud dropdown?

Choose Auto-detect. The current desktop picker exposes a subset of MAI-Transcribe-2’s supported languages; the model can automatically detect languages outside that manual list. Chinese (Traditional) is a separate output conversion from the Chinese transcript, rather than an additional model language.

Does my audio leave my computer?

With local transcription, speech recognition processes audio on your device. With cloud transcription, audio goes through the BridgeMind API to the hosted transcription provider. Optional Polish and Enhance Prompt send transcript text to hosted AI, even when the original transcription was local.

Is BridgeVoice included with BridgeMind Pro?

Yes. BridgeVoice is included with BridgeMind Pro. Cloud transcription is subject to usage limits. Optional Polish and Enhance Prompt use BridgeMind credits. See the pricing page for the current subscription and included credits.

Is the demo using my microphone?

No. This website demo uses prepared sample text and simulated recording states. It does not access your microphone, download a speech model, or call a transcription service. Download BridgeVoice to try dictation with your own voice.

Give your next idea a voice.

BridgeVoice is included with BridgeMind Pro. One subscription. More ways to build.

Download for macOSUniversal · DMG
Download for Windows64-bit Windows · Installer
Download for Linuxx86_64 · AppImage

Pro is $50/month or $480/year. See pricing and included credits.

BridgeVoice release notes