# Ubuntu 26.10 introduces Myna local speech-to-text tool for privacy-focused dictation

> **Key takeaway:** Ubuntu 26.10 features Myna, a local speech-to-text tool that utilizes models like Whisper or Parakeet to enable offline dictation.

**Published:** 2026-11-18T15:50:49+00:00 
**Source:** It&#x27;s FOSS Community News 
**Category:** distro-news 
**Canonical URL:** https://fosswire.org/news/ubuntu-2610-introduces-myna-local-speech-to-text-tool-for-pr.html 

## Summary
Canonical has integrated Myna into Ubuntu 26.10, providing users with a native speech-to-text utility. The tool allows users to dictate text directly into any active application via a keyboard shortcut, processing audio locally to ensure privacy.

Myna supports several backend speech models, including Parakeet, Whisper, and FunASR. By running these models on the local machine, the tool eliminates the need for internet connectivity once the initial model download is complete, preventing sensitive voice data from being sent to external servers.

## Technical implementation of local speech tools
- **Model Backend Support:** Myna utilizes local speech models such as Whisper, Parakeet, or FunASR to convert audio to text without cloud dependencies.
- **Transcription Pipeline:** Dictation workflows typically employ Voice Activity Detection (VAD), such as Silero VAD, to identify speech and trim silence before processing audio through the primary speech model.
- **Post-Processing via LLM:** Tools like Koe (written in Rust) can pass raw transcripts through a local LLM running in Ollama to perform punctuation correction and formatting cleanup.
- **Whisper Model Optimization:** For English-only dictation, using specialized '.en' models (e.g., small.en) provides higher accuracy and lower latency compared to multilingual versions of the same size.
- **Whisper-based File Support:** The WhisPaste 1.5 release adds local transcription capabilities for various audio and video formats, including MP3, FLAC, MKV, and WebM, using whisper.cpp.

## Why it matters
The rise of local speech-to-text tools like Myna and Koe demonstrates the viability of running heavy AI workloads on consumer-grade hardware. This shift enhances privacy for Linux users and reduces the reliance on proprietary, cloud-based dictation services.