Improve OpenMOSS audio tokenizer model card

README.md

@@ -4,7 +4,9 @@ language:
 - en
 license: apache-2.0
 library_name: mlx
-pipeline_tag: audio-to-audio
+pipeline_tag: feature-extraction
+base_model: OpenMOSS-Team/MOSS-Audio-Tokenizer
+base_model_relation: quantized
 tags:
 - mlx
 - audio
@@ -16,11 +18,11 @@ tags:
 - 8bit
 ---
 
-# OpenMOSS Audio Tokenizer — MLX
+# OpenMOSS Audio Tokenizer — MLX 8-bit
 
-The CAT codec component from the [OpenMOSS](https://github.com/open-moss) project, converted and quantized for native MLX inference on Apple Silicon.
+This repository contains an MLX-native int8 conversion of the OpenMOSS audio tokenizer for Apple Silicon.
 
-This is a supporting model — it encodes and decodes audio tokens for the MOSS TTS model family. It is not a standalone TTS model.
+It is a supporting model that encodes and decodes audio tokens for the OpenMOSS TTS family. It is not a standalone speech generation model.
 
 ## Variants
 
@@ -28,9 +30,17 @@ This is a supporting model — it encodes and decodes audio tokens for the MOSS
 | --- | --- |
 | `mlx-int8/` | int8 quantized weights |
 
+## Model Details
+
+- Developed by: AppAutomaton
+- Shared by: AppAutomaton on Hugging Face
+- Upstream model: [`OpenMOSS-Team/MOSS-Audio-Tokenizer`](https://huggingface.co/OpenMOSS-Team/MOSS-Audio-Tokenizer)
+- Task: audio tokenization and codec decoding
+- Runtime: MLX on Apple Silicon
+
 ## How to Get Started
 
-Load via [mlx-speech](https://github.com/appautomaton/mlx-speech):
+Load it directly with [`mlx-speech`](https://github.com/appautomaton/mlx-speech):
 
 ```python
 from mlx_speech.models.moss_audio_tokenizer import MossAudioTokenizerModel
@@ -38,7 +48,7 @@ from mlx_speech.models.moss_audio_tokenizer import MossAudioTokenizerModel
 model = MossAudioTokenizerModel.from_path("mlx-int8")
 ```
 
-The tokenizer is loaded automatically when you run any MOSS TTS generation script. You typically do not need to load it directly.
+The tokenizer is loaded automatically when you run OpenMOSS generation scripts. You usually do not need to instantiate it directly.
 
 ```bash
 python scripts/generate_moss_local.py \
@@ -46,12 +56,17 @@ python scripts/generate_moss_local.py \
   --output outputs/out.wav
 ```
 
-## Model Details
+## Notes
+
+- This repo contains the quantized MLX runtime artifact only.
+- The conversion remaps the original OpenMOSS audio tokenizer weights explicitly for MLX inference.
+- The artifact is shared by the OpenMOSS local TTS, TTSD, and SoundEffect runtime paths in this repo.
 
-Converted from the original OpenMOSS checkpoint using explicit MLX weight remapping — no PyTorch at inference time. Quantized to int8 with `W8Abf16` mixed precision.
+## Links
 
-See [mlx-speech](https://github.com/appautomaton/mlx-speech) for the full conversion pipeline and runtime code.
+- Source code: [mlx-speech](https://github.com/appautomaton/mlx-speech)
+- More examples: [AppAutomaton](https://github.com/appautomaton)
 
 ## License
 
-Apache 2.0 — following the upstream [OpenMOSS](https://github.com/open-moss) license terms.
+Apache 2.0 — following the upstream license published with [`OpenMOSS-Team/MOSS-Audio-Tokenizer`](https://huggingface.co/OpenMOSS-Team/MOSS-Audio-Tokenizer).