Breeze ASR 26 ONNX
The only build that runs on Android, iOS, and in the browser (WASM). Encoder and decoder are single ONNX files each. Slower on CPU (RTF 1.3) than CT2/GGML — use this only where you need mobile or WASM; otherwise...
Source du modèle
- Type
- Other
- Taille du fichier
- 730–961 MB
- Modèle de base
- MediaTek-Research/Breeze-ASR-26
- Version
- vRevision 17c563a
- Mis à jour
- 26 oct. 2026
- Téléchargements
- 0
- Source
- Hugging Face
Description de la source
Breeze-ASR-26 — ONNX (sherpa-onnx, INT8)
The only build that runs on Android, iOS, and in the browser (WASM). Encoder and decoder are single ONNX files each. Slower on CPU (RTF 1.3) than CT2/GGML — use this only where you need mobile or WASM; otherwise prefer CT2 (speed) or GGML (low RAM).
Part of the Breeze-ASR-26 edge family — the same MediaTek model in every runtime, pick by your constraint:
Repo Runtime RSS RTF (CPU 4-thread) Best for Breeze-ASR-26-ct2 CTranslate2 / faster-whisper ~2.9 GB 0.21 servers, 8 GB+ hosts, GPU Breeze-ASR-26-GGML whisper.cpp / MacWhisper 1.85 GB 0.40 4 GB hosts, desktop apps Breeze-ASR-26-ONNX sherpa-onnx / onnxruntime — 1.3 Android / iOS / WASM All Apache-2.0, derived from MediaTek-Research/Breeze-ASR-26. Measured on real multi-speaker Mandarin meeting audio. Mandarin does not regress; Taigi is transcribed as Mandarin meaning (not verbatim Taigi characters).
Usage
import sherpa_onnx
rec = sherpa_onnx.OfflineRecognizer.from_whisper(
encoder="breeze-asr-26-large-v2-encoder.int8.onnx",
decoder="breeze-asr-26-large-v2-decoder.int8.onnx",
tokens="breeze-asr-26-large-v2-tokens.txt",
language="zh", task="transcribe", num_threads=4)
Sources
1 source
Sources
1 sourceVérifié 26 oct.
Download files
Artefacts du modèle
2 artefacts
Artefacts du modèle
2 artefactsbreeze-asr-26-large-v2-decoder.int8.onnx
onnx · 961 MB · SHA-256 4e9bf6e78302…63de · Hugging Face
Download filebreeze-asr-26-large-v2-encoder.int8.onnx
onnx · 730 MB · SHA-256 53edc1ca0408…a9d4 · Hugging Face
Download fileExtraits de sources
3 extraits
Extraits de sources
3 extraits--- license: apache-2.0 base_model: MediaTek-Research/Breeze-ASR-26 language: [zh, nan] pipeline_tag: automatic-speech-recognition tags: [whisper, onnx, sherpa-onnx, taiwanese-hokkien, taigi, mobile, wasm, android, ios, int8, speech-recognition] --- # Breeze-ASR-26 — ONNX (sherpa-onnx, INT8) The **only** build that runs on **Android, iOS, and in the browser (WASM)**. Encoder and decoder are single ONNX files each. Slower on CPU (RTF 1.3) than CT2/GGML — use this only where you need mobile or WASM; otherwise prefer CT2 (speed) or GGML (low RAM). > Part of the **Breeze-ASR-26 edge family** — the same MediaTek model in every runtime, pick by your constraint: > > | Repo | Runtime | RSS | RTF (CPU 4-thread) | Best for | > |---|---|---|---|---| > | [Breeze-ASR-26-ct2](https://huggingface.co/weemed/Breeze-ASR-26-ct2) | CTranslate2 / faster-whisper | ~2.9 GB | **0.21** | servers, 8 GB+ hosts, GPU | > | [Breeze-ASR-26-GGML](https://huggingface.co/weemed/Breeze-ASR-26-GGML) | whisper.cpp / MacWhisper | **1.85 GB** | 0.40 | 4 GB hosts, desktop apps | > | [Breeze-ASR-26-ONNX](https://huggingface.co/weemed/Breeze-ASR-26-ONNX) | sherpa-onnx / onnxruntime | — | 1.3 | Android / iOS / WASM | > > All Apache-2.0, derived from [MediaTek-Research/Breeze-ASR-26](https://huggingface.co/MediaTek-Research/Breeze-ASR-26). Measured on real multi-speaker Mandarin meeting audio. Mandarin does not regress; Taigi is transcribed as Mandarin meaning (not verbatim Taigi characters). ## Usage ```python import sherpa_onnx rec = sherpa_onnx.OfflineRecognizer.from_whisper( encoder="breeze-asr-26-large-v2-encoder.int8.onnx", decoder="breeze-asr-26-large-v2-decoder.int8.onnx", tokens="breeze-asr-26-large-v2-tokens.txt", language="zh", task="transcribe", num_threads=4) ```
Source context: 0 downloads · 1 likes · Pipeline automatic-speech-recognition · Repo weemed/Breeze-ASR-26-ONNX
Source context: 0 downloads · 0 likes · Pipeline automatic-speech-recognition · Repo weemed/Breeze-ASR-26-ONNX
