Update main.py

main.py

@@ -5,6 +5,7 @@ import asyncio
 import aiohttp
 import tempfile
 import uuid
+import shutil
 import os
 import random
 import traceback
@@ -247,48 +248,107 @@ async def make_video(request: Request):
         # Generate a unique filename for the output
         output_filename = f"{uuid.uuid4()}.mp4"
         output_path = os.path.join(OUTPUT_DIR, output_filename)
+        temp_dir = os.path.join(tempfile.gettempdir(), generate_hash())
+        os.makedirs(temp_dir, exist_ok=True)
 
         data = await request.json()
-        image_urls = data.get("image_urls")
-        audio_url = data.get("audio_url")
-        duration = data.get("duration", 5)  # Default duration of 5 seconds per image
-
-        if not image_urls or not isinstance(image_urls, list) or not audio_url:
-            raise HTTPException(status_code=400, detail="Invalid image_urls or audio_url in request.")
-
-        # Download audio file
-        audio_file = await download_file(audio_url, ".mp3")
-
-        # Download image files and create a temporary file with the list of input files
-        image_files = []
-        concat_list_path = os.path.join(OUTPUT_DIR, "concat_list.txt")
-        with open(concat_list_path, "w") as f:
-            for i, url in enumerate(image_urls):
-                image_file = await download_file(url, f"_{i}.jpg")
-                image_files.append(image_file)
-                f.write(f"file '{image_file}'\nduration {duration}\n")  # Set duration for each image
-
-            # Add the last image again to ensure the audio doesn't cut off abruptly
-            f.write(f"file '{image_files[-1]}'\n")
-
-        # Run ffmpeg command to create video from images and add audio
-        ffmpeg_cmd = f"ffmpeg -f concat -safe 0 -i {concat_list_path} -i {audio_file} -c:v libx264 -pix_fmt yuv420p -c:a aac -shortest {output_path}"
-        process = await asyncio.create_subprocess_shell(
-            ffmpeg_cmd,
+        assets = data.get("assets", {})
+        clips = assets.get("clips", [])
+        music_url = assets.get("music_url")
+        volume_adjustment = data.get("volume_adjustment", 1.0)  # Default to normal volume if not specified
+
+        if not clips or not isinstance(clips, list):
+            raise HTTPException(status_code=400, detail="Invalid clips in request.")
+
+        # Create a list to hold clip video files that we'll concatenate later
+        clip_videos = []
+        segment_list_path = os.path.join(temp_dir, "segment_list.txt")
+        
+        with open(segment_list_path, "w") as segment_file:
+            # Process each clip: combine image with its audio
+            for i, clip in enumerate(clips):
+                image_url = clip.get("image_url")
+                audio_url = clip.get("audio_url")
+                
+                if not image_url or not audio_url:
+                    raise HTTPException(status_code=400, detail=f"Missing image_url or audio_url in clip {i}")
+                
+                # Download files
+                image_file = await download_file(image_url, ".jpg")
+                audio_file = await download_file(audio_url, ".mp3")
+                
+                # Create segment video
+                segment_output = os.path.join(temp_dir, f"segment_{i}.mp4")
+                
+                # Run ffmpeg command to create video segment
+                segment_cmd = f'ffmpeg -loop 1 -i {image_file} -i {audio_file} -c:v libx264 -tune stillimage -c:a aac -b:a 192k -shortest -pix_fmt yuv420p {segment_output}'
+                proc = await asyncio.create_subprocess_shell(
+                    segment_cmd,
+                    stdout=asyncio.subprocess.PIPE,
+                    stderr=asyncio.subprocess.PIPE
+                )
+                _, stderr = await proc.communicate()
+                
+                if proc.returncode != 0:
+                    print(f"FFmpeg error for segment {i}: {stderr.decode()}")
+                    raise HTTPException(status_code=500, detail=f"FFmpeg failed for segment {i}: {stderr.decode()}")
+                
+                # Add to list for concatenation
+                clip_videos.append(segment_output)
+                segment_file.write(f"file '{segment_output}'\n")
+                
+                # Clean up individual files
+                os.remove(image_file)
+                os.remove(audio_file)
+        
+        # Concatenate all segment videos
+        concat_output = os.path.join(temp_dir, "concat_output.mp4")
+        concat_cmd = f'ffmpeg -f concat -safe 0 -i {segment_list_path} -c copy {concat_output}'
+        
+        proc = await asyncio.create_subprocess_shell(
+            concat_cmd,
             stdout=asyncio.subprocess.PIPE,
             stderr=asyncio.subprocess.PIPE
         )
-        stdout, stderr = await process.communicate()
-
-        if process.returncode != 0:
-            print(f"FFmpeg error: {stderr.decode()}")
-            raise HTTPException(status_code=500, detail=f"FFmpeg failed: {stderr.decode()}")
-
-        # Clean up temporary files
-        for file in image_files:
-            os.remove(file)
-        os.remove(audio_file)
-        os.remove(concat_list_path)
+        _, stderr = await proc.communicate()
+        
+        if proc.returncode != 0:
+            print(f"FFmpeg concat error: {stderr.decode()}")
+            raise HTTPException(status_code=500, detail=f"FFmpeg concat failed: {stderr.decode()}")
+        
+        # If there's a music URL, download it and mix with the video
+        if music_url:
+            music_file = await download_file(music_url, ".wav")
+            
+            # Final command to mix music with video at specified volume
+            final_cmd = (
+                f'ffmpeg -i {concat_output} -i {music_file} -filter_complex '
+                f'"[1:a]volume={volume_adjustment}[music];[0:a][music]amix=inputs=2:duration=longest" '
+                f'-c:v copy {output_path}'
+            )
+            
+            proc = await asyncio.create_subprocess_shell(
+                final_cmd,
+                stdout=asyncio.subprocess.PIPE,
+                stderr=asyncio.subprocess.PIPE
+            )
+            _, stderr = await proc.communicate()
+            
+            if proc.returncode != 0:
+                print(f"FFmpeg final mix error: {stderr.decode()}")
+                raise HTTPException(status_code=500, detail=f"FFmpeg music mixing failed: {stderr.decode()}")
+            
+            os.remove(music_file)
+        else:
+            # If no music, just copy the concatenated output
+            shutil.copy(concat_output, output_path)
+        
+        # Clean up temporary files and directory
+        for video_file in clip_videos:
+            os.remove(video_file)
+        os.remove(segment_list_path)
+        os.remove(concat_output)
+        os.rmdir(temp_dir)
 
         # Return the URL path to the output file
         return f"/output/{output_filename}"