Tojito
Linux desktop MVP
Local AI for the sound on your computer.
Tojito is a working Linux desktop MVP — real desktop audio → speech recognition → captions → translation → GNOME overlay has been validated. Built by DPSonic. Windows support is next.
Why local
Your audio stays with you.
Tojito runs on your machine. The goal is AI help for desktop audio without sending your streams, calls, or conversations through someone else’s servers. Models are managed by the app, and core processing stays local as the product grows.
- On-device processingReal-time audio work without a cloud audio pipeline.
- Offline-capable directionLocal models and on-device workflows as the platform matures.
- Active projectPublic updates track what is shipping and what is still ahead.
Available now
Built and validated
What works today
Linux desktop MVP
Working tray app on Linux: real desktop audio through local speech recognition, captions, translation, and a GNOME overlay — validated on Ubuntu GNOME Wayland.
Real-time local audio processing
Process desktop audio streams on-device. Your audio stays on your machine.
Source captions
On-device speech recognition with live source captions on the GNOME overlay.
Offline speech translation
Local caption translation on the validated Linux path (including EN↔RU acceptance). No cloud round-trip for this audio path.
GNOME overlay product UX
Desktop overlay for captions and translation with live opacity and enable/disable controls — acceptance-backed on the Shell-extension path.
Engineering progress
Measured acceptance lives in Updates
M11 first-user productization on Ubuntu 24.04 GNOME Wayland — evidence over marketing. Read the measured update.
Coming next
Direction — not shipping yet
Broader platform ideas under development. These are goals, not available features.
Windows / WASAPI
Windows desktop capture and end-to-end product path. Not available yet.
Verified profanity mute on the product ASR path
Filter ON mute is not verified yet on the real Pack ASR path. Wiring exists; live mute acceptance remains open. Default remains off.
Real-time voice translation & TTS
Spoken voice translation and speech synthesis as the engine matures.
Semantic content filtering
Policy-aware filtering beyond keyword and profanity matching.
Audio enhancement
Clarity, noise, and voice quality improvements powered by local models.
Additional languages & policies
More languages, dialects, and configurable filtering policies.
Cross-platform MVP & private beta
Broader platform packaging and a private beta once the Windows path lands.
How it is built
Desktop app. Local models. Your control.
Tojito is a desktop product. When builds are ready, the app connects to your audio and manages the models it needs. This site is the public presence and development journal — not a place to hand-manage model files.
- Process computer audioWork with the sound coming from your desktop environment.
- Keep it localPrefer on-device inference and local model management.
- Evolve capabilitiesCaptions and translation on Linux today; Windows and broader platform work next.
- Publish progressDevelopment updates form a searchable public history of the product.