main

app.py

@@ -1,76 +1,34 @@
 import gradio as gr
-import os
 import subprocess
-from ttv import generate_video  # We'll modify ttv.py to make it a function
+import os
 
-def install_dependencies():
+def run_ttv():
     try:
-        # Cài đặt diffusers từ git
-        subprocess.run(["pip", "install", "-U", "git+https://github.com/huggingface/diffusers"], 
-                      check=True,
-                      capture_output=True)
+        result = subprocess.run(['python', 'ttv.py'], capture_output=True, text=True)
         
-        # Cài đặt inference-script
-        subprocess.run(["pip", "install", "-e", ".[inference-script]"], 
-                      check=True,
-                      capture_output=True)
-        
-        return "✅ Đã cài đặt thành công các gói phụ thuộc!"
-    except subprocess.CalledProcessError as e:
-        return f"❌ Lỗi khi cài đặt: {str(e)}"
-
-def text_to_video(prompt, negative_prompt):
-    # Generate video from text
-    output_path = generate_video(prompt, negative_prompt)
-    return output_path
+        # Lấy danh sách file mp3 trong thư mục temp
+        mp3_files = [f for f in os.listdir('temp') if f.endswith('.mp3')]
+        if mp3_files:
+            # Trả về đường dẫn đến file mp3 mới nhất
+            latest_mp3 = os.path.join('temp', mp3_files[-1])
+            return result.stdout, latest_mp3
+        return result.stdout, None
+    except Exception as e:
+        return f"Có lỗi xảy ra: {str(e)}", None
 
-def list_videos():
-    # List all MP4 files in the current directory
-    videos = [f for f in os.listdir('.') if f.endswith('.mp4')]
-    return videos
-
-# Create Gradio interface
+# Tạo giao diện
 with gr.Blocks() as demo:
-    gr.Markdown("# Chuyển Văn Bản Thành Video")
-    
-    # Add installation button at the top
-    install_btn = gr.Button("Cài đặt Dependencies")
-    install_output = gr.Textbox(label="Trạng thái cài đặt", interactive=False)
+    gr.Markdown("# Ứng dụng Text-to-Voice")
     
     with gr.Row():
-        with gr.Column():
-            # Input components
-            text_input = gr.Textbox(label="Nhập nội dung video", lines=3)
-            neg_prompt = gr.Textbox(label="Từ khóa loại trừ", 
-                                  value="chất lượng kém, chuyển động không đồng nhất, mờ, giật, biến dạng")
-            generate_btn = gr.Button("Tạo Video")
-        
-        with gr.Column():
-            # Output video display
-            video_output = gr.Video(label="Video đã tạo")
-    
-    # Gallery of existing videos
-    gr.Markdown("### Video đã tạo trước đó")
-    gallery = gr.Gallery(
-        value=list_videos(), 
-        label="Video có sẵn", 
-        show_label=True, 
-        elem_id="gallery",
-        columns=2
-    )
+        run_button = gr.Button("Chạy Text-to-Voice")
+        output = gr.Textbox(label="Kết quả")
     
-    # Connect components
-    generate_btn.click(
-        fn=text_to_video,
-        inputs=[text_input, neg_prompt],
-        outputs=[video_output]
-    )
+    # Thêm audio player
+    audio_output = gr.Audio(label="File âm thanh đã tạo")
     
-    # Connect install button
-    install_btn.click(
-        fn=install_dependencies,
-        inputs=[],
-        outputs=[install_output]
-    )
+    run_button.click(fn=run_ttv, outputs=[output, audio_output])
 
-demo.launch()
+# Chạy ứng dụng
+if __name__ == "__main__":
+    demo.launch()

requirements.txt

@@ -1,4 +1,4 @@
-diffusers
-torch
-transformers
-numpy
+diffusers>=0.30.3
+transformers>=0.44.2
+accelerate>=0.34.0
+imageio-ffmpeg>=0.5.1

ttv.py

@@ -1,21 +1,26 @@
 import torch
-from diffusers import LTXPipeline
-from diffusers.utils import export_to_video
-import os
+from diffusers import CogVideoXImageToVideoPipeline
+from diffusers.utils import export_to_video, load_image
 
-def generate_video(prompt, negative_prompt):
-    pipe = LTXPipeline.from_pretrained("Lightricks/LTX-Video", torch_dtype=torch.bfloat16)
-    pipe.to("cuda")
-    
-    video = pipe(
-        prompt=prompt,
-        negative_prompt=negative_prompt,
-        width=704,
-        height=480,
-        num_frames=161,
-        num_inference_steps=50, 
-    ).frames[0]
-    
-    output_path = f"output_{len(os.listdir('.'))}.mp4"
-    export_to_video(video, output_path, fps=24)
-    return output_path
+prompt = "A little girl is riding a bicycle at high speed. Focused, detailed, realistic."
+image = load_image(image="input.jpg")
+pipe = CogVideoXImageToVideoPipeline.from_pretrained(
+    "THUDM/CogVideoX-5b-I2V",
+    torch_dtype=torch.bfloat16
+)
+
+pipe.enable_sequential_cpu_offload()
+pipe.vae.enable_tiling()
+pipe.vae.enable_slicing()
+
+video = pipe(
+    prompt=prompt,
+    image=image,
+    num_videos_per_prompt=1,
+    num_inference_steps=50,
+    num_frames=49,
+    guidance_scale=6,
+    generator=torch.Generator(device="cuda").manual_seed(42),
+).frames[0]
+
+export_to_video(video, "temp/output.mp4", fps=8)