Update app.py

app.py

@@ -1,12 +1,16 @@
 import streamlit as st
 import os
 import tempfile
-from moviepy.editor import ImageSequenceClip
+from moviepy.editor import ImageSequenceClip, concatenate_videoclips
 from PIL import Image
 import torch
 from diffusers import AudioLDMPipeline
 from transformers import AutoProcessor, ClapModel, BlipProcessor, BlipForConditionalGeneration
 
+# Charger le modèle et le processeur Blip
+processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-base")
+model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-base")
+
 # make Space compatible with CPU duplicates
 if torch.cuda.is_available():
     device = "cuda"
@@ -22,14 +26,10 @@ pipe.unet = torch.compile(pipe.unet)
 
 # CLAP model (only required for automatic scoring)
 clap_model = ClapModel.from_pretrained("sanchit-gandhi/clap-htsat-unfused-m-full").to(device)
-processor = AutoProcessor.from_pretrained("sanchit-gandhi/clap-htsat-unfused-m-full")
+audio_processor = AutoProcessor.from_pretrained("sanchit-gandhi/clap-htsat-unfused-m-full")
 
 generator = torch.Generator(device)
 
-# Charger le modèle et le processeur Blip pour la description d'images
-image_caption_processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-base")
-image_caption_model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-base")
-
 # Streamlit app setup
 st.set_page_config(
     page_title="Text to Media",
@@ -55,12 +55,12 @@ if uploaded_files:
             f.write(uploaded_file.read())
         image_paths.append(image_path)
         
-        # Générez la légende pour chaque image
+        # Générer la légende pour chaque image
         try:
             image = Image.open(image_path).convert("RGB")
-            inputs = image_caption_processor(image, return_tensors="pt")
-            out = image_caption_model.generate(**inputs)
-            caption = image_caption_processor.decode(out[0], skip_special_tokens=True)
+            inputs = processor(image, return_tensors="pt")
+            out = model.generate(**inputs)
+            caption = processor.decode(out[0], skip_special_tokens=True)
             descriptions.append(caption)
         except Exception as e:
             descriptions.append("Erreur lors de la génération de la légende")
@@ -79,7 +79,7 @@ if uploaded_files:
     n_candidates = st.slider("Number waveforms to generate", 1, 3, 3, 1)
 
     def score_waveforms(text, waveforms):
-        inputs = processor(text=text, audios=list(waveforms), return_tensors="pt", padding=True)
+        inputs = audio_processor(text=text, audios=list(waveforms), return_tensors="pt", padding=True)
         inputs = {key: inputs[key].to(device) for key in inputs}
         with torch.no_grad():
             logits_per_text = clap_model(**inputs).logits_per_text  # this is the audio-text similarity score