voice-to-text that runs entirely on your device. Zero cloud, zero subscriptions. Qwen3-ASR engine

Hey everyone! 👋

I built Brethof Voice Pro because I was frustrated with two things: voice-to-text tools that only work well in English, and the fact that every good STT solution either costs $700 (Dragon) or requires sending your voice to the cloud every month.

Everything runs 100% offline. Your voice data never leaves your machine. Not "mostly local" — zero network calls during transcription.

We use the Qwen3-ASR engine, which benchmarks at 1.84% word error rate across 10 languages — that's significantly better than Whisper. The numbers are published in arXiv paper 2601.21337.

Hardware flexibility was key. GGUF models with Vulkan acceleration work on NVIDIA, AMD, and Intel GPUs — plus CPU-only mode. 6 model tiers from 1 GB to 3.2 GB.

36 languages with auto-detection. Thai, Vietnamese, Polish, Arabic, Cantonese, Korean, Turkish — languages that have been an afterthought in every other STT product. Hotword dictionaries for domain terms, and LoRA voice training is coming.

$49 one-time. No subscription. 14-day free trial, no credit card.

Would love your feedback. What languages would you test first?