Deploy HARP wrapper via model agent

.harp/manifest.json

@@ -0,0 +1,18 @@
+{
+  "entry": "app.py",
+  "framework": "voicefixer",
+  "generated": true,
+  "io": {
+    "inputs": [
+      "audio",
+      "dropdown"
+    ],
+    "outputs": [
+      "audio"
+    ]
+  },
+  "repo_id": "haoheliu/voicefixer",
+  "source": "recipe",
+  "space_layout": "huggingface-gradio",
+  "task": "declipping"
+}

README.md

@@ -1,13 +1,19 @@
 ---
-title: Voicefixer
-emoji: 🏢
-colorFrom: red
-colorTo: green
+title: "VoiceFixer"
+colorFrom: indigo
+colorTo: gray
 sdk: gradio
-sdk_version: 6.20.0
-python_version: '3.13'
+sdk_version: 5.28.0
 app_file: app.py
 pinned: false
+license: "mit"
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+# VoiceFixer
+
+VoiceFixer aims to restore human speech regardless how serious its degraded. It can handle noise, reverberation, low resolution (2kHz~44.1kHz) and clipping (0.1-1.0 threshold) effect within one model.
+
+- Inputs: audio, dropdown
+- Outputs: audio
+
+Generated by the HARP model agent from a recipe.

app.py

@@ -0,0 +1,61 @@
+from __future__ import annotations
+
+import gradio as gr
+try:
+    import spaces
+except ImportError:  # 'spaces' is only provided by Hugging Face Spaces
+    import types as _types
+
+    def _gpu(*args, **kwargs):
+        if len(args) == 1 and callable(args[0]) and not kwargs:
+            return args[0]
+
+        def _decorator(func):
+            return func
+
+        return _decorator
+
+    spaces = _types.SimpleNamespace(GPU=_gpu)
+
+from pyharp import *
+
+
+import tempfile
+from voicefixer import VoiceFixer
+
+# Initialize VoiceFixer. It handles downloading checkpoints and setting device.
+# The VoiceFixer class automatically detects and uses CUDA if available.
+voicefixer_model = VoiceFixer()
+
+
+model_card = ModelCard(
+    name="VoiceFixer",
+    description="VoiceFixer aims to restore human speech regardless how serious its degraded. It can handle noise, reverberation, low resolution (2kHz~44.1kHz) and clipping (0.1-1.0 threshold) effect within one model.",
+    author="haoheliu",
+    tags=["declipping", "denoise", "dereverberation", "mel", "speech", "speech-analysis", "speech-enhancement", "speech-processing", "speech-synthesis", "super-resolution", "tts", "vocoder"],
+)
+
+
+@spaces.GPU
+def process_fn(input_audio, mode):
+    output_file = tempfile.NamedTemporaryFile(suffix=".wav", delete=False).name
+    voicefixer_model.restore(input_audio, output_file, mode=int(mode))
+    return output_file
+
+
+with gr.Blocks() as demo:
+    input_components = [
+        gr.Audio(type="filepath", label="Input Audio").harp_required(True).set_info("Upload an audio file to be processed by VoiceFixer."),
+        gr.Dropdown(choices=["0", "1", "2"], value="0", label="Processing Mode", info="Select the VoiceFixer processing mode:\n0: Original Model (suggested by default)\n1: Add preprocessing module (remove higher frequency)\n2: Train mode (might work sometimes on seriously degraded real speech)"),
+    ]
+    output_components = [
+        gr.Audio(type="filepath", label="Fixed Audio").set_info("The enhanced audio output from VoiceFixer."),
+    ]
+    build_endpoint(
+        model_card=model_card,
+        input_components=input_components,
+        output_components=output_components,
+        process_fn=process_fn,
+    )
+
+demo.queue().launch(share=True, show_error=False, pwa=True)

packages.txt

@@ -0,0 +1 @@
+ffmpeg

requirements.txt

@@ -0,0 +1,6 @@
+git+https://github.com/TEAMuP-dev/pyharp.git@v0.3.0
+gradio>=4.0
+git+https://github.com/haoheliu/voicefixer.git@main
+torch>=1.7.0
+librosa
+torchlibrosa