Improve model card: add pipeline tag and project page

This PR significantly improves the model card for SonicMaster by:
- Adding the `pipeline_tag: audio-to-audio`, ensuring the model can be found easily under the audio pipeline (e.g., at https://huggingface.co/models?pipeline_tag=audio-to-audio).
- Expanding the model description with key information from the paper abstract, detailing its purpose, control mechanisms, and approach.
- Including a link to the official project page for further details and potential usage examples.

README.md

@@ -1,7 +1,16 @@
 ---
 license: apache-2.0
+pipeline_tag: audio-to-audio
 ---
 
-TBA soon
+# SonicMaster: Towards Controllable All-in-One Music Restoration and Mastering
 
-[Read paper](https://huggingface.co/papers/2508.03448)
+[Read paper](https://huggingface.co/papers/2508.03448) | [Project Page](https://amaai-lab.github.io/SonicMaster/)
+
+SonicMaster is the first unified generative model for music restoration and mastering that addresses a broad spectrum of audio artifacts with text-based control. This model can be conditioned on natural language instructions to apply targeted enhancements, or it can operate in an automatic mode for general restoration.
+
+The approach leverages a flow-matching generative training paradigm to learn an audio transformation that maps degraded inputs to their cleaned, mastered versions guided by text prompts. Objective audio quality metrics and subjective listening tests demonstrate that SonicMaster significantly improves sound quality across various artifact categories, confirming its effectiveness.
+
+## Usage
+
+TBA soon. Please refer to the [project page](https://amaai-lab.github.io/SonicMaster/) for current usage instructions and code examples.