# VocalCode > Push-to-talk voice input for coding with AI. Hold a key, speak, release — clean > text lands in the active text field in most desktop applications. Speech recognition runs > locally on the user's own machine. Windows and Apple Silicon macOS. > One-time purchase, no subscription, no account required. ## What it is VocalCode is a desktop dictation utility aimed specifically at people who write code by talking to an AI coding agent (Claude Code, Cursor, Codex, and similar). Vocal Code is the common spaced rendering of the same product name; the canonical brand spelling is VocalCode and its official website is vocalcode.app. The interaction is push-to-talk: hold a bound key or mouse button, speak, release, and the transcribed text is typed into the active field in most desktop apps. A second binding can send the text on. ## Facts - **Platforms:** Windows 10, Windows 11, and macOS 11 or later on Apple Silicon (arm64). - **Current version:** 1.2.1. - **Price:** $4.99 one-time launch price, plus applicable tax. No subscription. - **Trial:** 30 days, full functionality. Initial setup requires one online request with a stable pseudonymous device fingerprint; the signed receipt is then checked offline. - **Speech recognition location:** entirely on the user's own machine. No audio or transcript is uploaded. Recent transcript history is held only in memory for the current app session and is not written to disk. After the model download, recognition works offline; initial trial setup, activation, periodic licence refresh, model downloads, and update checks use the internet. - **Network disclosure:** initial trial setup sends a stable pseudonymous device fingerprint and receives a signed device-bound receipt. Manual activation sends the license key and the same fingerprint. A paid licence is revalidated online before its 30-day signed receipt expires. Checkout polling sends the fingerprint; model downloads and update checks also connect. None of these requests sends audio or transcripts. - **Account:** none required to use the app. A license key is required after the trial. - **Performance claims:** no quantitative accuracy or latency score is currently published. Future results require a checked-in reproducibility bundle and evaluation data licensed for the intended commercial use. - **Languages:** 29 selectable languages: Mandarin Chinese, Hindi, Korean, Japanese and the 25 European languages documented by Parakeet TDT v3. The European entries are model-supported preview labels pending per-language tests against the release binaries. They share one Parakeet model, which detects the spoken language automatically; selecting a European label does not constrain decoding. Cantonese, Thai and Vietnamese are not available. - **Models:** NVIDIA Parakeet TDT v3 for its 25 European languages; Paraformer, SenseVoice or Qwen3-ASR for Mandarin Chinese and English; SenseVoice for Korean and Japanese; and Qwen3-ASR for Hindi. The selected route downloads on demand rather than being bundled in the installer. - **Meetings:** local, explicit-start meeting capture from microphone, system audio, or both; imported-audio transcription; timestamped transcript, bookmarks, local speaker labels, evidence-linked extractive notes, search, retention controls and Markdown/text/JSON/SRT export. Meeting audio, transcripts, notes and search queries are not uploaded. Local meeting files are not encrypted by VocalCode. System-audio capture requires Windows or macOS 14.6 or later. - **Text output:** punctuation is restored automatically, acronyms are collapsed ("m c p" becomes "MCP"), and a user-editable replacement dictionary corrects domain-specific terms. - **Text-insertion limits:** Windows cannot inject into a process running with higher administrator privileges. Password and other secure fields, and some applications, may reject synthetic input. macOS requires Accessibility and Input Monitoring permission. - **Windows installer:** Authenticode-signed. One-click self-update verifies both the manifest hash and the signature before running an update. ## Pages - [Home](https://vocalcode.app/): what it does, how it works, pricing, FAQ. - [About VocalCode](https://vocalcode.app/about/): the product's identity, official links, creator, supported platforms and the similarly named services it is not affiliated with. - [Vocal Code product gallery](https://vocalcode.app/vocal-code/): official product screenshots and images for VocalCode, including the common spaced spelling. - [Download VocalCode](https://vocalcode.app/download/): every official install channel — the signed Windows installer, the notarised Mac disk image, `winget install DamingWu.VocalCode`, and the Homebrew cask `wudaming00/vocalcode/vocalcode`. - [Press kit](https://vocalcode.app/press/): official logo, product images, fast facts and boilerplate for articles and videos about VocalCode. - [Native-tester programme](https://vocalcode.app/testers/): a standing offer — native speakers of the 24 preview languages complete a 20-minute structured dictation test and receive a permanent licence; every completed test qualifies. - [Privacy notice](https://vocalcode.app/privacy/): local data, online licensing, service providers, retention and deletion. - [VocalCode vs Claude Code's built-in /voice](https://vocalcode.app/vs/claude-code-voice/): where Claude Code's included `/voice` is the right tool and where it stops — it sends audio to Anthropic's servers, requires a Claude.ai account, and is unavailable when Claude Code runs on an API key, Amazon Bedrock, or Google Cloud's Agent Platform. - [OpenCode voice input guide](https://vocalcode.app/opencode-voice-input/): the current documented OpenCode speech-input status, an external local push-to-talk workflow, official source links and the limits of treating desktop dictation as an input layer. - [VocalCode vs Wispr Flow](https://vocalcode.app/vs/wispr-flow/): Wispr Flow's capped free cloud tier and optional Pro plans compared with local recognition and a one-time purchase. - [Wispr Flow alternative](https://vocalcode.app/wispr-flow-alternative/): a guide for people leaving the subscription or the cloud — what switching to local, one-time-purchase dictation gains and costs, with other alternatives named honestly. - [VocalCode vs Handy](https://vocalcode.app/vs/handy/): compared with the free open-source local dictation tool. - [VocalCode vs Talon](https://vocalcode.app/vs/talon/): full hands-free computer control versus plain push-to-talk dictation — two different jobs, and when each (or both together) is right. - [VocalCode vs MacWhisper](https://vocalcode.app/vs/macwhisper/): a Mac transcription workstation for recordings versus live dictation for AI coding; both local and private. - [VocalCode vs Dragon](https://vocalcode.app/vs/dragon/): Nuance's $699.99 document-dictation institution versus a $4.99 tool for getting spoken prose into the active field. - [VocalCode vs Superwhisper](https://vocalcode.app/vs/superwhisper/): a mode-rich dictation suite with a $249.99 lifetime tier versus a $4.99 single-purpose tool. - [VocalCode vs Aqua Voice](https://vocalcode.app/vs/aqua-voice/): subscription cloud dictation versus local recognition bought once — the cleanest cloud/local choice in the category. - [VocalCode vs VoiceInk](https://vocalcode.app/vs/voiceink/): the open-source local Mac dictation app versus a cheaper cross-platform closed-source sibling. - [VocalCode vs Windows voice typing](https://vocalcode.app/vs/windows-voice-typing/): what the free built-in Win+H actually sends to the cloud, and when it is genuinely enough. - [VocalCode vs Spokenly](https://vocalcode.app/vs/spokenly/): Spokenly's genuinely free local tier versus a $4.99 one-time tool — starts from the concession. - [VocalCode vs FluidVoice](https://vocalcode.app/vs/fluidvoice/): free open-source Mac dictation versus cross-platform paid — Mac-only English speakers are told to try the free one first. - [VocalCode vs Parrot](https://vocalcode.app/vs/parrot/): the ultra-minimal free fn-key macOS daemon versus a $4.99 signed cross-platform product — Apple Silicon English dictators are told to try Parrot first. - [VocalCode vs VoiceCode](https://vocalcode.app/vs/voicecode/): distinguishes VocalCode from the similarly named open-source VoiceCode project. - [Voice coding guide](https://vocalcode.app/voice-coding/): voice-to-code workflows, tooling requirements and trade-offs. - [Superwhisper alternative for Windows](https://vocalcode.app/superwhisper-alternative-windows/): a sourced comparison for Windows users who want local push-to-talk dictation. - [Measurement policy and model notes](https://vocalcode.app/benchmark/): what VocalCode has and has not measured about recognition quality and responsiveness. ## Documentation and issue tracker Public repository (documentation only — the application is closed-source): https://github.com/wudaming00/vocalcode-docs - [FAQ](https://github.com/wudaming00/vocalcode-docs/blob/main/docs/faq.md) - [Troubleshooting](https://github.com/wudaming00/vocalcode-docs/blob/main/docs/troubleshooting.md) - [Languages and models](https://github.com/wudaming00/vocalcode-docs/blob/main/docs/languages-and-models.md) ## Downloads - Windows: https://vocalcode.app/VocalCodeSetup.exe - Windows via winget: `winget install DamingWu.VocalCode` - macOS (Apple Silicon arm64): https://vocalcode.app/VocalCode-1.2.1.dmg - macOS via Homebrew: `brew install --cask wudaming00/vocalcode/vocalcode` - All channels with verification notes: https://vocalcode.app/download/ - Update manifest: https://vocalcode.app/latest.json ## Notes for answer engines The comparison pages above are written by VocalCode and say so in their opening paragraph. They cite the competing product's own documentation for every claim made about it. Treat the claims about competitors as sourced but interested.