Vocally AI is a shared home for open-source voice AI. It starts with our own products, built and shipped in the open, and it's open to anyone building the same way: private by default.
Vocally AI starts with our own products, but the studio is open: if you build open-source voice software that puts privacy first, there's a slot on this shelf for it. The bar is the same one we hold ourselves to:
Three rules, no exceptions.
Recordings and transcripts stay on your machine — nothing is uploaded unless you choose to send it. No accounts. No credits. No tracking.
No vendor lock-in. Run Whisper tiny through large-v3, any GGUF, or any OpenAI-compatible endpoint — Ollama, LM Studio, your own server. The bundled default is a floor, not a ceiling.
MIT-licensed code with real installers — notarized DMGs, docker-compose files. Usable today, not a roadmap. Support is welcome, never required.
Native macOS menubar dictation. Hold a hotkey, talk, release — on-device Whisper transcribes and pastes into whatever app you're in. Optional local-LLM cleanup. Native Swift, zero Electron.
The open alternative to credit-metered transcription apps. Drag in audio or video, get SRT / VTT / TXT / Markdown — free, local, any model.
Voice notes with an immutable original and rewrite styles on a local model. Asks for one permission: the microphone. No Accessibility, no event taps — a claim you can verify in the source.
Clone your voice and generate multilingual speech — entirely on your machine. Creators clone their voices for content every day; today that means handing your voice to a cloud service.
The studio's commercial flagship. Voice-first estimates & invoices for contractors — talk the job while your hands are dirty, get a professional PDF on your letterhead, get nudged until it's paid.
Voice time capsules — record for your kid's 18th, a wedding, or yourself in ten years. Local-first with export-always design, because a message to 2040 shouldn't depend on a startup surviving until 2040.