ray-amjad/hyperwhisper-app

HyperWhisper — open-source speech-to-text for macOS, Windows & iOS. Local transcription and bring-your-own-key is free and open; HyperWhisper Cloud is a paid OpenRouter-like router.

C#

56

879 commits

updated Oct 5, 2026

See the code

README

HyperWhisper

Fast, private speech-to-text for macOS, Windows, and Linux.

Press a key, speak, and your words appear as text in any app — transcribed on-device by default, with an optional hosted Cloud for the highest-accuracy models.

Created by Ray Amjad.

Website · Docs · Download · YouTube

License: Apache 2.0


What this is

HyperWhisper is a dictation app: a global hotkey records your voice, transcribes it (locally via Whisper / Parakeet, or via HyperWhisper Cloud), optionally post-processes it with an LLM, and pastes the result wherever your cursor is.

This repository is the full source for the apps, the Fly.io transcription backend, the marketing/license website, the documentation, and the shared schemas — open-sourced under Apache-2.0.

Fully open source — including the backend

Everything in HyperWhisper is open source under Apache-2.0 — the macOS, Windows, Linux, and iOS apps, the marketing/license website, the docs, the shared schemas, and the Fly.io transcription backend (hyperwhisper-cloud/). Nothing is closed source. Local transcription runs entirely on your machine with no account and no network — clone it, build it, use it.

HyperWhisper Cloud is the one paid piece — the managed, hosted instance we run for the largest, most accurate models. Its backend lives in this repo, so you can read, audit, and run it yourself; metering stays server-side, so building from source doesn't unlock our hosted Cloud for free. Everything else (local models, post-processing, modes, vocabulary, backup/restore) works with no account.

Cloud runs on prepaid credits — one balance for every provider. No per-provider API keys to manage, no subscription, no juggling separate bills: top up once and pay only for what you transcribe. Buy credits, activate your Account Key in the app, and you're done — it funds development of everything in this repo.

Repository layout

PathWhat it is
app/macosmacOS app (Swift / SwiftUI)
app/windowsWindows app (C# / WPF / .NET 10)
app/linuxLinux app (C# / Avalonia / .NET 10)
app/iosiOS app (work in progress; not built here)
hyperwhisper-cloudFly.io edge transcription service
nextjsMarketing & account website (Next.js)
mintlify-helpDocumentation site (Mintlify)
shared-core-rsRust shared core (UniFFI) used by the native apps
shared-modelsPer-model metadata catalog
shared-promptsPost-processing prompt templates
shared-app-classificationApp-type & cloud-STT catalogs
shared-backupCross-platform backup schema
shared-typesShared TypeScript types
integrationsExternal integrations (MCP, editor plugins)
toolsNative build scripts (parakeet-engine, sherpa-onnx)

This repo uses a git submodule for the backend:

git clone --recurse-submodules https://github.com/ray-amjad/hyperwhisper-app.git
# or, after a plain clone:
git submodule update --init --recursive

Building

Each app has its own setup; see the per-directory AGENTS.md for details.

  • macOS — open app/macos/hyperwhisper.xcodeproj in Xcode and run (⌘R). Dependencies resolve via Swift Package Manager. Requires macOS 14+.
  • Windows — dotnet run -c Debug from app/windows/HyperWhisper/. Requires the .NET 10 SDK on Windows. See app/windows/AGENTS.md for installer builds.
  • Linux — dotnet build app/linux/HyperWhisper.Linux/HyperWhisper.Linux.csproj -c Release. Requires the .NET 10 SDK. Run app/linux/scripts/run-ui-smoke.sh for the headless UI smoke test; see packaging/linux/README.md for Debian packaging, or download Linux v1.0.0.
  • Web — npm install && npm run dev in nextjs/. Local builds use SKIP_ENV_VALIDATION=1 npm run build.

Building official, signed/notarized releases additionally requires signing credentials and a Sentry DSN, injected from CI secrets — these are not needed to build and run from source.

Third-party components

See THIRD_PARTY_LICENSES.md for attribution of bundled and linked dependencies (most MIT / Apache-2.0; a few require attribution, e.g. the Silero VAD model under CC-BY-4.0).

Contributing

Issues and pull requests are welcome. Please keep changes focused and follow the conventions documented in the relevant AGENTS.md.

License

Apache License 2.0 — see LICENSE. The code for everything, including the backend, is covered by this license. The hosted HyperWhisper Cloud service — our managed instance and prepaid credits — is a separate paid offering, even though its backend source is included here under the same license.

The "HyperWhisper" name, logo, and app icon are not covered by the Apache license — see TRADEMARKS.md. You can fork and ship the code, but under your own name and branding, not as the official HyperWhisper.

ray-amjad/hyperwhisper-app

HyperWhisper — open-source speech-to-text for macOS, Windows & iOS. Local transcription and bring-your-own-key is free and open; HyperWhisper Cloud is a paid OpenRouter-like router.

C#

56

879 commits

updated Oct 5, 2026

See the code

README

HyperWhisper

Fast, private speech-to-text for macOS, Windows, and Linux.

Press a key, speak, and your words appear as text in any app — transcribed on-device by default, with an optional hosted Cloud for the highest-accuracy models.

Created by Ray Amjad.

Website · Docs · Download · YouTube

License: Apache 2.0


What this is

HyperWhisper is a dictation app: a global hotkey records your voice, transcribes it (locally via Whisper / Parakeet, or via HyperWhisper Cloud), optionally post-processes it with an LLM, and pastes the result wherever your cursor is.

This repository is the full source for the apps, the Fly.io transcription backend, the marketing/license website, the documentation, and the shared schemas — open-sourced under Apache-2.0.

Fully open source — including the backend

Everything in HyperWhisper is open source under Apache-2.0 — the macOS, Windows, Linux, and iOS apps, the marketing/license website, the docs, the shared schemas, and the Fly.io transcription backend (hyperwhisper-cloud/). Nothing is closed source. Local transcription runs entirely on your machine with no account and no network — clone it, build it, use it.

HyperWhisper Cloud is the one paid piece — the managed, hosted instance we run for the largest, most accurate models. Its backend lives in this repo, so you can read, audit, and run it yourself; metering stays server-side, so building from source doesn't unlock our hosted Cloud for free. Everything else (local models, post-processing, modes, vocabulary, backup/restore) works with no account.

Cloud runs on prepaid credits — one balance for every provider. No per-provider API keys to manage, no subscription, no juggling separate bills: top up once and pay only for what you transcribe. Buy credits, activate your Account Key in the app, and you're done — it funds development of everything in this repo.

Repository layout

PathWhat it is
app/macosmacOS app (Swift / SwiftUI)
app/windowsWindows app (C# / WPF / .NET 10)
app/linuxLinux app (C# / Avalonia / .NET 10)
app/iosiOS app (work in progress; not built here)
hyperwhisper-cloudFly.io edge transcription service
nextjsMarketing & account website (Next.js)
mintlify-helpDocumentation site (Mintlify)
shared-core-rsRust shared core (UniFFI) used by the native apps
shared-modelsPer-model metadata catalog
shared-promptsPost-processing prompt templates
shared-app-classificationApp-type & cloud-STT catalogs
shared-backupCross-platform backup schema
shared-typesShared TypeScript types
integrationsExternal integrations (MCP, editor plugins)
toolsNative build scripts (parakeet-engine, sherpa-onnx)

This repo uses a git submodule for the backend:

git clone --recurse-submodules https://github.com/ray-amjad/hyperwhisper-app.git
# or, after a plain clone:
git submodule update --init --recursive

Building

Each app has its own setup; see the per-directory AGENTS.md for details.

  • macOS — open app/macos/hyperwhisper.xcodeproj in Xcode and run (⌘R). Dependencies resolve via Swift Package Manager. Requires macOS 14+.
  • Windows — dotnet run -c Debug from app/windows/HyperWhisper/. Requires the .NET 10 SDK on Windows. See app/windows/AGENTS.md for installer builds.
  • Linux — dotnet build app/linux/HyperWhisper.Linux/HyperWhisper.Linux.csproj -c Release. Requires the .NET 10 SDK. Run app/linux/scripts/run-ui-smoke.sh for the headless UI smoke test; see packaging/linux/README.md for Debian packaging, or download Linux v1.0.0.
  • Web — npm install && npm run dev in nextjs/. Local builds use SKIP_ENV_VALIDATION=1 npm run build.

Building official, signed/notarized releases additionally requires signing credentials and a Sentry DSN, injected from CI secrets — these are not needed to build and run from source.

Third-party components

See THIRD_PARTY_LICENSES.md for attribution of bundled and linked dependencies (most MIT / Apache-2.0; a few require attribution, e.g. the Silero VAD model under CC-BY-4.0).

Contributing

Issues and pull requests are welcome. Please keep changes focused and follow the conventions documented in the relevant AGENTS.md.

License

Apache License 2.0 — see LICENSE. The code for everything, including the backend, is covered by this license. The hosted HyperWhisper Cloud service — our managed instance and prepaid credits — is a separate paid offering, even though its backend source is included here under the same license.

The "HyperWhisper" name, logo, and app icon are not covered by the Apache license — see TRADEMARKS.md. You can fork and ship the code, but under your own name and branding, not as the official HyperWhisper.