MIT-licensed voice typing for Mac, Windows, and Linux that works with Groq, Deepgram, AssemblyAI, OpenAI, Claude, Gemini, Ollama and more via your own API keys.
OpenTypeless is the most provider-agnostic app in this directory. Rather than picking a speech model and a language model for you, it lets you wire up almost any combination: Groq Whisper or Deepgram or AssemblyAI or Apple's built-in speech for transcription, then OpenAI or Claude or Gemini or a local Ollama model for the polish pass. It's MIT licensed and free if you bring your own keys, with optional paid cloud plans if you'd rather not.
At a glance
Platforms
macOS, Windows, Linux
Pricing
Free with your own API keys; optional hosted cloud plans
No vendor lock-in at all. Swap providers when pricing changes or a better model ships. Nothing else here gives you this much control over the stack.
More than dictation. "Ask Anything" does one-shot voice Q&A, selected-text editing rewrites what you've highlighted, and translation mode switches target languages on the fly.
App-aware writing. Adapts tone and structure to the application you're typing into — recently improved in v1.1.49.
Custom dictionary and local correction rules. Fixes recurring mistranscriptions deterministically rather than hoping the model learns.
Genuinely cross-platform and MIT. Three desktop platforms, permissive licence.
Where it falls short
Mostly cloud in practice. Despite being self-hosted software, the default configurations send audio to Groq, Deepgram, or OpenAI. Only Apple Speech and Ollama keep things local, and that pairing is the weakest on quality. The previous "offline" tag on this entry was misleading and has been corrected.
You are the systems integrator. Two API keys minimum, two billing relationships, and your own judgement about which models pair well. There is no sensible default that just works.
No independent reviews. At ~440 stars this is a small project and I found no third-party testing. Feature claims come from the repository.
Costs are unpredictable. Per-token pricing across two providers is harder to budget than a flat $10/month.
Small maintainer base. 24 open pull requests suggests review capacity is the bottleneck.
Pricing
Free under MIT if you supply your own API keys — you then pay Groq, OpenAI, and the rest directly, typically a few dollars a month for moderate use. Optional hosted cloud plans exist for people who don't want to manage keys.
Who it's for
Tinkerers who want to choose their own models and would rather pay providers at cost than pay a markup to a dictation vendor. If that description makes you tired, almost anything else in this category will suit you better — Handy for free and local, Wispr Flow for paid and polished.
Last verified: 9 August 2026. No independent reviews located; details are from the project's own repository.
Similar to OpenTypeless
OpenTypeless: Bring your own keys and mix any speech and language model you like – Dictation Database
Five platforms and an AppSumo lifetime deal, with the risks that implies
General Dictation
Cloud dictation across Mac, Windows, Linux, iPhone, and Android with zero-retention processing. Free tier is 1,000 words a month; Pro is $12/month annually.
