AVA
Persian, English and Russian speech-to-text for Windows. One floating orb, nothing else: tap it, speak, tap again — the text is already typed into whatever application you were writing in.
- Persian
- English
- Russian
- Auto language
- Offline
- Windows Download · 334 MB
- macOS Source build
- Linux Source build
- Android Source build
- iOS Source build
- iPadOS Source build
- HarmonyOS Scaffold
What it is
AVA is a desktop dictation app for Persian (fa-IR), English and Russian. It sits on the desktop
as a single floating orb, records locally, recognises speech on your own computer and types the result
straight into the focused window. No audio leaves the machine.
Automatic language detection, four models
Auto is the default card: it identifies the language on every phrase on this PC and sends the audio to the matching offline model.
| Model | Language | Size | Speed |
|---|---|---|---|
| Koochik (fast) | Persian | ~0.45 GB | about 5× faster than real time |
| Nemotron (accurate) | Persian | ~2.5 GB | slower than real time on a CPU |
| Parakeet (English) | English | ~670 MB | about 1 s per phrase |
| GigaAM (Russian) | Russian | ~225 MB | about 0.3 s per phrase |
In Auto the English model is a smaller Parakeet 110M (~465 MB, ~0.1 s per phrase) and the detector itself is Whisper base, used only to choose the language and never to transcribe. Models are downloaded once, on first use, and cached locally; after that AVA works fully offline.
Privacy & local processing
- 100% Offline. All audio processing and transcription are executed entirely on your local machine.
- Total Privacy. Your voice data and text are never transmitted to any cloud servers or third-party APIs.
- Volatile Memory. Audio chunks are processed in RAM and discarded immediately after typing.
- One network call. The only outbound request is the one-time model download from Hugging Face.
Third-party licenses
AVA runs fully offline and incorporates the following open-weights models and libraries. Model weights are downloaded on first use, not bundled, and keep their own licences. Application code is MIT.
- Nemotron ASR Streaming Farsi (Persian, accurate) — mehdi-hf/nemotron-asr-streaming-farsi · OpenMDW License Agreement, v1.1
- Shenava-Koochik v1.0 (Persian, fast CTC) — Apache License 2.0 (apache.org)
- NVIDIA Parakeet (English) — parakeet-tdt-0.6b-v3 · CC-BY-4.0
- GigaAM v3 (Russian) — ai-sage/GigaAM-v3 · MIT
- OpenAI Whisper base (language detector only) — openai/whisper-base · MIT
- sherpa-onnx — k2-fsa/sherpa-onnx offline speech engine
AVA — Your Voice. Your Words.
Built & Engineered by DANISHYARSI