BitVoice ASR models
On-device speech-recognition models for the BitVoice dictation app, run via
CRISP ASR (dart:ffi + ggml).
BitVoice has no in-app networking: download a model here in your browser, then
import the file into the app (Model manager, Import). Each model is a single
.gguf or .bin; Moonshine models are a .zip (model + tokenizer) that the
app unpacks on import.
75 models, 15 architectures, 28 languages. See SOURCES.md for the upstream of each model.
Recommended (balanced starter set)
| Model | Lang | Size | Quant | Download |
|---|---|---|---|---|
| SenseVoice small 路 multilingual | English, Chinese, Japanese, Korean, Cantonese | 129 MB | q4_k | sensevoice-small-q4.gguf |
| Whisper small 路 multilingual (99 lang) | Multilingual | 252 MB | q8_0 | whisper-small-q8.bin |
| Whisper Hindi+English (Hinglish) 路 base | Hindi, English | 78 MB | q8_0 | whisper-hindi-swift-q8.bin |
| Parakeet 110M 路 English | English | 75 MB | q4_k | parakeet-110m-en-q4.gguf |
| Whisper distil-small 路 English | English | 175 MB | q8_0 | whisper-distil-small-en-q8.bin |
| Paraformer 路 Chinese | Chinese | 123 MB | q4_k | paraformer-zh-q4.gguf |
| Kotoba-Whisper v2.0 路 Japanese | Japanese | 513 MB | q5_0 | whisper-kotoba-ja-q5.bin |
| FastConformer CTC 路 Spanish | Spanish | 112 MB | q8_0 | fastconformer-es-q8.gguf |
All models and languages
| Model | Lang | Size | Quant | Download |
|---|---|---|---|---|
| Whisper base 路 English | English | 57 MB | q5_1 | whisper-base-en-q5.bin |
| Moonshine tiny 路 English | English | 19 MB | q4 | moonshine-tiny-en-q4.zip |
| wav2vec2 XLSR 路 English | English | 212 MB | q4_k | wav2vec2-xlsr-en-q4.gguf |
| GigaAM v3 路 Russian | Russian | 237 MB | q8_0 | gigaam-v3-ru-q8.gguf |
| Nemotron 3.5 路 English (streaming) | English | 390 MB | q4_k | nemotron-streaming-en-q4.gguf |
| Parakeet 0.6B v3 路 English | English | 399 MB | q4_k | parakeet-0.6b-en-q4.gguf |
| Qwen3-ASR 0.6B 路 multilingual | English, Chinese, Multilingual | 602 MB | q4_k | qwen3-asr-q4.gguf |
| Canary 1B v2 路 multilingual | English, German, Spanish, French | 582 MB | q4_k | canary-1b-q4.gguf |
| ReazonSpeech v2 路 Japanese | Japanese | 636 MB | q8_0 | reazonspeech-ja-q8.gguf |
| Whisper Hindi+English (Hinglish) 路 large | Hindi, English | 452 MB | q4_0 | whisper-hindi-apex-q4.bin |
| Whisper tiny 路 multilingual (99 lang) | Multilingual | 42 MB | q8_0 | whisper-tiny-q8.bin |
| Whisper base 路 multilingual (99 lang) | Multilingual | 78 MB | q8_0 | whisper-base-q8.bin |
| wav2vec2 XLSR 路 French | French | 356 MB | q8_0 | wav2vec2-fr-q8.gguf |
| wav2vec2 XLSR 路 Spanish | Spanish | 356 MB | q8_0 | wav2vec2-es-q8.gguf |
| wav2vec2 XLSR 路 Italian | Italian | 356 MB | q8_0 | wav2vec2-it-q8.gguf |
| wav2vec2 XLSR 路 Japanese | Japanese | 365 MB | q8_0 | wav2vec2-ja-q8.gguf |
| wav2vec2 XLSR 路 Chinese | Chinese | 370 MB | q8_0 | wav2vec2-zh-q8.gguf |
| wav2vec2 XLSR 路 Portuguese | Portuguese | 356 MB | q8_0 | wav2vec2-pt-q8.gguf |
| wav2vec2 XLSR 路 Dutch | Dutch | 356 MB | q8_0 | wav2vec2-nl-q8.gguf |
| wav2vec2 XLSR 路 Arabic | Arabic | 356 MB | q8_0 | wav2vec2-ar-q8.gguf |
| data2vec audio 路 English | English | 119 MB | q8_0 | data2vec-en-q8.gguf |
| HuBERT large 路 English | English | 212 MB | q4_k | hubert-en-q4.gguf |
| omniASR CTC 300M 路 multilingual | Multilingual | 343 MB | q8_0 | omniasr-ctc-300m-q8.gguf |
| Kyutai STT 1B 路 English/French | English, French | 636 MB | q4_k | kyutai-stt-enfr-q4.gguf |
| Moonshine base 路 English | English | 69 MB | q8_0 | moonshine-base-en-q8.zip |
| Moonshine base 路 German | German | 36 MB | q4_k | moonshine-base-de-q4.zip |
| Moonshine tiny 路 Japanese | Japanese | 15 MB | q4_k | moonshine-tiny-ja-q4.zip |
| Moonshine tiny 路 Korean | Korean | 15 MB | q4_k | moonshine-tiny-ko-q4.zip |
| Moonshine tiny 路 Chinese | Chinese | 16 MB | q4_k | moonshine-tiny-zh-q4.zip |
| Moonshine tiny 路 Arabic | Arabic | 16 MB | q4_k | moonshine-tiny-ar-q4.zip |
| FastConformer CTC 路 Arabic | Arabic | 112 MB | q8_0 | fastconformer-ar-q8.gguf |
| FastConformer CTC 路 Belarusian | Belarusian | 112 MB | q8_0 | fastconformer-be-q8.gguf |
| FastConformer CTC 路 German | German | 112 MB | q8_0 | fastconformer-de-q8.gguf |
| FastConformer CTC 路 English | English | 118 MB | q8_0 | fastconformer-en-q8.gguf |
| FastConformer CTC 路 Persian | Persian | 112 MB | q8_0 | fastconformer-fa-q8.gguf |
| FastConformer CTC 路 French | French | 112 MB | q8_0 | fastconformer-fr-q8.gguf |
| FastConformer CTC 路 Croatian | Croatian | 112 MB | q8_0 | fastconformer-hr-q8.gguf |
| FastConformer CTC 路 Armenian | Armenian | 112 MB | q8_0 | fastconformer-hy-q8.gguf |
| FastConformer CTC 路 Italian | Italian | 112 MB | q8_0 | fastconformer-it-q8.gguf |
| FastConformer CTC 路 Georgian | Georgian | 112 MB | q8_0 | fastconformer-ka-q8.gguf |
| FastConformer CTC 路 Kazakh/Russian | Kazakh, Russian | 112 MB | q8_0 | fastconformer-kk-ru-q8.gguf |
| FastConformer CTC 路 Dutch | Dutch | 112 MB | q8_0 | fastconformer-nl-q8.gguf |
| FastConformer CTC 路 Polish | Polish | 112 MB | q8_0 | fastconformer-pl-q8.gguf |
| FastConformer CTC 路 Russian | Russian | 112 MB | q8_0 | fastconformer-ru-q8.gguf |
| FastConformer CTC 路 Ukrainian | Ukrainian | 112 MB | q8_0 | fastconformer-ua-q8.gguf |
| FastConformer CTC 路 Uzbek | Uzbek | 112 MB | q8_0 | fastconformer-uz-q8.gguf |
| Parakeet TDT 0.6B 路 Japanese | Japanese | 642 MB | q8_0 | parakeet-ja-q8.gguf |
| Parakeet 路 German | German | 643 MB | q8_0 | parakeet-de-q8.gguf |
| FireRedASR2 AED 路 Chinese | Chinese | 918 MB | q4_k | firered-zh-q4.gguf |
| Moonshine tiny 路 Vietnamese | Vietnamese | 16 MB | q4_k | moonshine-tiny-vi-q4.zip |
| Moonshine tiny 路 Ukrainian | Ukrainian | 16 MB | q4_k | moonshine-tiny-uk-q4.zip |
| Parakeet TDT 0.6B v2 路 English | English | 379 MB | q4_k | parakeet-tdt-v2-en-q4.gguf |
| Parakeet RNNT 0.6B 路 English | English | 379 MB | q4_k | parakeet-rnnt-en-q4.gguf |
| Parakeet Unified 0.6B 路 English | English | 380 MB | q4_k | parakeet-unified-en-q4.gguf |
| Parakeet TDT-CTC 1.1B 路 English | English | 654 MB | q4_k | parakeet-tdt-ctc-1.1b-en-q4.gguf |
| Parakeet CTC 1.1B 路 Japanese | Japanese | 640 MB | q4_k | parakeet-ctc-1.1b-ja-q4.gguf |
| FastConformer CTC xlarge 路 English | English | 366 MB | q4_k | fastconformer-en-xlarge-q4.gguf |
| omniASR CTC 1B 路 multilingual | Multilingual | 129 MB | q4_k | omniasr-ctc-1b-q4.gguf |
| wav2vec2 XLS-R 路 Czech | Czech | 212 MB | q4_k | wav2vec2-cs-q4.gguf |
| wav2vec2 XLS-R 1B TEVR 路 German | German | 577 MB | q4_k | wav2vec2-de-tevr-q4.gguf |
| Whisper distil-large-v3 路 English | English | 513 MB | q5_0 | whisper-distil-large-v3-en-q5.bin |
| Whisper distil-medium 路 English | English | 410 MB | q8_0 | whisper-distil-medium-en-q8.bin |
| Whisper small 路 Tamil | Tamil | 252 MB | q8_0 | whisper-tamil-small-q8.bin |
| Kotoba-Whisper bilingual 路 Japanese/English | Japanese, English | 513 MB | q5_0 | whisper-kotoba-jaen-q5.bin |
| Whisper large-v3-turbo 路 German | German | 547 MB | q5_0 | whisper-turbo-de-q5.bin |
| Whisper medium 路 Urdu | Urdu | 559 MB | q5_1 | whisper-urdu-medium-q5.bin |
| Whisper large-v3-turbo 路 Korean | Korean | 595 MB | q5_1 | whisper-korean-turbo-q5.bin |
Notes:
- CTC models (wav2vec2, data2vec, HuBERT, omniASR, FastConformer CTC) emit lowercase text without punctuation.
- Larger models (1B+ Parakeet/Canary/Qwen3/Kyutai/ReazonSpeech, Whisper medium/large) are best on a high-RAM phone (about 6 GB or more).
Provenance and attribution
Re-hosted quantized conversions. Most GGUF conversions are by
cstr; Whisper .bin files come from
ggerganov/whisper.cpp and
credible community converters (distil-whisper, kotoba-tech, Pomni). The two
Hindi+English (Hinglish) models and the re-quantized Urdu and Korean models are
BitVoice's own conversions. See SOURCES.md for per-model upstreams.
CC-BY-4.0 models (Parakeet, Canary, Kyutai) are redistributed with attribution;
consult each original model card for authoritative terms.
- Downloads last month
- 235
Hardware compatibility
Log In to add your hardware
We're not able to determine the quantization variants.
Inference Providers NEW
This model isn't deployed by any Inference Provider. 馃檵 Ask for provider support