Update app.py

app.py

@@ -3,8 +3,8 @@ import torch
 from diffusers import DiffusionPipeline, DPMSolverMultistepScheduler
 from diffusers.utils import export_to_video
 import uuid
-import os
 
+# Model Yükleme
 model_id = "vdo/zeroscope_v2_576w"
 pipe = DiffusionPipeline.from_pretrained(model_id, torch_dtype=torch.float32)
 pipe.scheduler = DPMSolverMultistepScheduler.from_config(pipe.scheduler.config)
@@ -12,23 +12,25 @@ pipe.to("cpu")
 
 def generate_video(prompt):
     try:
-        # KALİTEYİ ZİRVEYE TAŞIDIK (CPU limitlerini zorluyoruz)
-        # num_inference_steps: 25 (Görüntü çamurdan kurtulacak)
-        # num_frames: 24 (Video 3 saniye ve daha akıcı olacak)
+        # Kaliteyi koruyalım (Patronun istediği gibi net olsun)
+        # num_inference_steps=20 (Dengeli kalite)
         frames = pipe(
             prompt, 
-            num_inference_steps=25, 
+            num_inference_steps=20, 
             height=320, 
             width=576, 
-            num_frames=24
+            num_frames=16
         ).frames
         
-        unique_name = f"viral_video_{uuid.uuid4()}.mp4"
-        export_to_video(frames[0], unique_name, fps=8)
-        return os.path.abspath(unique_name)
+        output_filename = f"viral_{uuid.uuid4()}.mp4"
+        export_to_video(frames[0], output_filename, fps=8)
+        
+        # SADECE dosya adını döndür, Gradio bunu otomatik URL'ye çevirir
+        return output_filename
     except Exception as e:
-        print(f"HATA: {e}")
+        print(f"HATA: {str(e)}")
         return None
 
+# API İsmi: predict
 demo = gr.Interface(fn=generate_video, inputs="text", outputs="video", api_name="predict")
 demo.launch()