Return to gradio 5.28.0 with pyharp v0.3.0: this stack cannot take huggingface-hub 1.x

README.md

@@ -4,7 +4,7 @@ emoji: 🗣️
 colorFrom: indigo
 colorTo: blue
 sdk: gradio
-sdk_version: 6.24.0
+sdk_version: 5.28.0
 app_file: app.py
 pinned: false
 license: other

app.py

@@ -9,7 +9,6 @@ from pathlib import Path
 import gradio as gr
 import spaces
 from pyharp import ModelCard, build_endpoint
-from pyharp.tags import Subcategory
 
 ROOT = Path(__file__).resolve().parent
 sys.path.append(str(ROOT / "indextts"))
@@ -53,40 +52,36 @@ def synthesize(text, voice_reference, emotion_reference, emotion_weight):
         raise gr.Error(f"IndexTTS2 synthesis failed: {exc}") from exc
 
 
-if __name__ == "__main__":
-    with gr.Blocks(title="IndexTTS2") as demo:
-        input_components = [
-            gr.Textbox(label="Text to Synthesize", lines=3).set_info(
-                "Chinese or English text to synthesize."
-            ),
-            gr.Audio(type="filepath", label="Voice Reference").set_info(
-                "Required voice sample; the model uses up to 15 seconds."
-            ),
-            gr.Audio(type="filepath", label="Emotion Reference")
-            .harp_required(False)
-            .set_info("Optional. Leave empty to follow the voice reference."),
-            gr.Slider(0.0, 1.0, value=0.8, step=0.05, label="Emotion Weight").set_info(
-                "Strength of the optional emotion reference."
-            ),
-        ]
-        output_components = [
-            gr.Audio(type="filepath", label="Output Audio").set_info("Generated WAV audio."),
-        ]
-        build_endpoint(
-            model_card=ModelCard(
-                name="IndexTTS2",
-                description=(
-                    "Zero-shot speech synthesis with a voice reference and optional emotion "
-                    "reference. Repository: https://github.com/index-tts/index-tts Paper: "
-                    "https://arxiv.org/abs/2601.03888"
-                ),
-                author="Yunpei Li, Xun Zhou, Jinchao Wang, et al.",
-                tags=[Subcategory.TEXT_TO_SPEECH, 'text-to-speech', 'voice-cloning', 'emotion-control'],
-            ),
-            input_components=input_components,
-            output_components=output_components,
-            process_fn=synthesize,
-        )
+with gr.Blocks(title="IndexTTS2") as demo:
+    input_components = [
+        gr.Textbox(label="Text to Synthesize", lines=3).set_info(
+            "Chinese or English text to synthesize."
+        ),
+        gr.Audio(type="filepath", label="Voice Reference").set_info(
+            "Required voice sample; the model uses up to 15 seconds."
+        ),
+        gr.Audio(type="filepath", label="Emotion Reference")
+        .harp_required(False)
+        .set_info("Optional. Leave empty to follow the voice reference."),
+        gr.Slider(0.0, 1.0, value=0.8, step=0.05, label="Emotion Weight").set_info(
+            "Strength of the optional emotion reference."
+        ),
+    ]
+    output_components = [
+        gr.Audio(type="filepath", label="Output Audio").set_info("Generated WAV audio."),
+    ]
+    build_endpoint(
+        model_card=ModelCard(
+            name="IndexTTS2",
+            description="Zero-shot speech synthesis with a voice reference and optional emotion reference.",
+            author="IndexTeam",
+            tags=["text-to-speech", "voice-cloning", "emotion-control"],
+        ),
+        input_components=input_components,
+        output_components=output_components,
+        process_fn=synthesize,
+    )
 
 
-    demo.queue(default_concurrency_limit=1).launch(share=True, show_error=True, pwa=True)
+if __name__ == "__main__":
+    demo.queue(default_concurrency_limit=1).launch(show_error=True, pwa=True)

requirements.txt

@@ -1,7 +1,7 @@
 accelerate==1.8.1
 descript-audiotools==0.7.2
-transformers==5.17.0
-tokenizers
+transformers==4.52.4
+tokenizers==0.21.0
 cn2an==0.5.22
 ffmpeg-python==0.2.0
 Cython==3.0.7
@@ -17,7 +17,7 @@ munch==4.0.0
 opencv-python==4.9.0.80
 tensorboard==2.9.1
 librosa==0.10.2.post1
-safetensors
+safetensors==0.5.2
 deepspeed==0.17.1
 modelscope==1.27.0
 omegaconf
@@ -31,4 +31,4 @@ torchaudio==2.8.0
 
 WeTextProcessing; platform_machine != "Darwin"
 wetext; platform_system == "Darwin"
-git+https://github.com/TEAMuP-dev/pyharp.git@v0.4.0
+git+https://github.com/TEAMuP-dev/pyharp.git@v0.3.0