Breeze ASR 26 Ct2
The fastest edge build: RTF 0.21 on CPU, real-time with headroom. Use this on servers, 8 GB+ hosts, or GPU. For 4 GB hosts or desktop apps use the GGML repo; for mobile/WASM use ONNX.
Model source
- Kind
- Other
- File size
- 1.45 GB
- Base model
- MediaTek-Research/Breeze-ASR-26
- Version
- vRevision b68b37a
- Updated
- Oct 26, 2026
- Downloads
- 51
- Source
- Hugging Face
Source description
Breeze-ASR-26 — CTranslate2 / faster-whisper (INT8)
The fastest edge build: RTF 0.21 on CPU, real-time with headroom. Use this on servers, 8 GB+ hosts, or GPU. For 4 GB hosts or desktop apps use the GGML repo; for mobile/WASM use ONNX.
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
from faster_whisper import WhisperModel
m = WhisperModel("weemed/Breeze-ASR-26-ct2", device="cpu", compute_type="int8", cpu_threads=4)
segments, _ = m.transcribe("meeting.wav", language="zh", beam_size=1)
print("".join(s.text for s in segments))
INT8 vs FP32: CER 4.69% (function-word/segmentation jitter only), 2.4x faster, quarter the size.
Sources
1 source
Sources
1 sourceVerified Oct 26
Model artifacts
1 artifact
Model artifacts
1 artifactSource excerpts
3 excerpts
Source excerpts
3 excerpts--- license: apache-2.0 base_model: MediaTek-Research/Breeze-ASR-26 language: [zh, nan] pipeline_tag: automatic-speech-recognition library_name: ctranslate2 tags: [whisper, taiwanese-hokkien, taigi, ctranslate2, faster-whisper, int8, real-time, speech-recognition] --- # Breeze-ASR-26 — CTranslate2 / faster-whisper (INT8) The **fastest** edge build: RTF 0.21 on CPU, real-time with headroom. Use this on servers, 8 GB+ hosts, or GPU. For 4 GB hosts or desktop apps use the GGML repo; for mobile/WASM use ONNX. > 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 from faster_whisper import WhisperModel m = WhisperModel("weemed/Breeze-ASR-26-ct2", device="cpu", compute_type="int8", cpu_threads=4) segments, _ = m.transcribe("meeting.wav", language="zh", beam_size=1) print("".join(s.text for s in segments)) ``` INT8 vs FP32: CER 4.69% (function-word/segmentation jitter only), 2.4x faster, quarter the size.
Source context: 51 downloads · 0 likes · Pipeline automatic-speech-recognition · Library ctranslate2 · Repo weemed/Breeze-ASR-26-ct2
Source context: 16 downloads · 0 likes · Pipeline automatic-speech-recognition · Library ctranslate2 · Repo weemed/Breeze-ASR-26-ct2
