Downloads · 30 days
18
20% of all-time downloads
funasr-nano-coreml
korakotlee/funasr-nano-coreml
funasr-nano-coreml is a automatic speech recognition model from korakotlee. Use it when you need speech turned into text. It is set up for coreml. The card lists the license as other.
CoreML conversion of FunAudioLLM/Fun-ASR-Nano-2512 for on-device inference on Apple Silicon, optimized for KoMeet Desktop.
All-time downloads
88
Public
Repo size
3 GB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.bin3.3 GB · 100%
At a glance
- Task
- Automatic Speech Recognition
- Library
- coreml
- License
- other
- Model type
- funasr-nano
- Access
- Public
- Created
- Sep 9, 2026
- Updated
- Sep 10, 2026
- SHA
- d578acce
Model card
From the Hugging Face model README
Fun-ASR-Nano - CoreML (Apple Silicon / Neural Engine)
CoreML conversion of FunAudioLLM/Fun-ASR-Nano-2512 for on-device inference on Apple Silicon, optimized for KoMeet Desktop.
Combines a SANM acoustic feature encoder with a Qwen3 autoregressive language model decoder.
Files
| File | Precision | Role |
|---|---|---|
FunASREncoder.mlmodelc | FLOAT16 | SANM acoustic feature encoder |
FunASRDecoder.mlmodelc | FLOAT16 | Qwen3 language model decoder |
config.json | - | Model configuration and hyper-parameters |
vocab.json | - | Token vocabulary mapping |
komeet-manifest.json | - | KoMeet asset integrity manifest |
License & Attribution
Weights derived from FunAudioLLM/Fun-ASR-Nano-2512.
- Task
- Automatic Speech Recognition
- Library
- coreml
- Type
- funasr-nano
- License
- other
- Languages
- zh, en, ja, ko, yue
- Created
- Sep 9, 2026
- Updated
- Sep 10, 2026