Update app.py

app.py

@@ -7,16 +7,7 @@ lpmc_client = gr.load("seungheondoh/LP-Music-Caps-demo", src="spaces")
 from gradio_client import Client
 
 client = Client("https://fffiloni-test-llama-api.hf.space/", hf_token=hf_token)
-
-from diffusers import DiffusionPipeline
-import torch
-
-pipe = DiffusionPipeline.from_pretrained("stabilityai/stable-diffusion-xl-base-1.0", torch_dtype=torch.float16, use_safetensors=True, variant="fp16")
-pipe.to("cuda")
-#pipe.enable_model_cpu_offload()
-
-# if using torch < 2.0
-# pipe.enable_xformers_memory_efficient_attention()
+zrscp_client = Client("https://fffiloni-zeroscope--76p9h.hf.space/", hf_token=hf_token)
 
 from pydub import AudioSegment
 
@@ -30,9 +21,12 @@ def cut_audio(input_path, output_path, max_duration=30000):
 
     return output_path
 
-def solo_xd(prompt):
-    images = pipe(prompt=prompt).images[0]
-    return images
+def solo_zrscp(prompt):
+    res_vid = zrscp_client.predict(
+        prompt,
+        api_name="/zrscp"
+    )
+    return res_vid
 
 def infer(audio_file):
 
@@ -82,12 +76,15 @@ def infer(audio_file):
     
     print(f"Llama2 result: {result}")
 
-    images = pipe(prompt=result).images[0]
+    res_vid = zrscp_client.predict(
+        prompt,
+        api_name="/zrscp"
+    )
 
     print("Finished")
     
     #return cap_result, result, images
-    return images, result, gr.update(visible=True)
+    return res_vid, result, gr.update(visible=True)
 
 css = """
 #col-container {max-width: 510px; margin-left: auto; margin-right: auto;}
@@ -118,18 +115,18 @@ with gr.Blocks(css=css) as demo:
         infer_btn = gr.Button("Generate Image from Music")
         #lpmc_cap = gr.Textbox(label="Lp Music Caps caption")
         llama_trans_cap = gr.Textbox(label="Llama translation", visible=False)
-        img_result = gr.Image(label="Image Result")
+        vid_result = gr.Video(label="Image Result")
         tryagain_btn = gr.Button("Try again ?", visible=False)
 
         gr.Examples(examples=[["./examples/electronic.mp3"],["./examples/folk.wav"], ["./examples/orchestra.wav"]],
                     fn=infer,
                     inputs=[audio_input],
-                    outputs=[img_result, llama_trans_cap, tryagain_btn],
+                    outputs=[vid_result, llama_trans_cap, tryagain_btn],
                     cache_examples=True
                    )
 
     #infer_btn.click(fn=infer, inputs=[audio_input], outputs=[lpmc_cap, llama_trans_cap, img_result])
-    infer_btn.click(fn=infer, inputs=[audio_input], outputs=[img_result, llama_trans_cap, tryagain_btn])
-    tryagain_btn.click(fn=solo_xd, inputs=[llama_trans_cap], outputs=[img_result])
+    infer_btn.click(fn=infer, inputs=[audio_input], outputs=[vid_result, llama_trans_cap, tryagain_btn])
+    tryagain_btn.click(fn=solo_zrscp, inputs=[llama_trans_cap], outputs=[vid_result])
 
 demo.queue(max_size=20).launch()