Upload app.py with huggingface_hub

app.py

@@ -62,7 +62,7 @@ def transcribe(ref_audio_path):
 
 
 @spaces.GPU(duration=120)
-def synthesize(text, ref_audio_path, ref_text, speaker_id, max_length_ms, temperature, topk):
+def synthesize(text, ref_audio_path, ref_text, speaker_id, max_length_s, temperature, topk):
     text = (text or "").strip()
     if not text:
         raise gr.Error("Please enter some text to synthesize.")
@@ -92,7 +92,7 @@ def synthesize(text, ref_audio_path, ref_text, speaker_id, max_length_ms, temper
         text=text,
         speaker=int(speaker_id),
         context=context,
-        max_audio_length_ms=float(max_length_ms),
+        max_audio_length_ms=float(max_length_s) * 1000.0,
         temperature=float(temperature),
         topk=int(topk),
     )
@@ -120,7 +120,7 @@ with gr.Blocks(title="Miso TTS 8B") as demo:
                 )
             with gr.Accordion("Advanced", open=False):
                 speaker_id = gr.Slider(0, 1, value=0, step=1, label="Speaker ID")
-                max_length = gr.Slider(2000, 30000, value=10000, step=1000, label="Max audio length (ms)")
+                max_length = gr.Slider(2, 60, value=10, step=1, label="Max audio length (s)")
                 temperature = gr.Slider(
                     0.1, 1.5, value=0.9, step=0.05,
                     label="Temperature (auto-lowered when cloning a voice)",