スレッドセーフに修正、動画ソート順不具合修正

app.py

@@ -48,6 +48,7 @@ class Replicate:
         self.output_url = None
         self.response = None
         self.prediction_id = None
+        self.lock = threading.Lock()
 
     def run_replicate(self, retries=0):
         try:
@@ -128,9 +129,10 @@ class Replicate:
             print(traceback.format_exc())
             
     def download_and_save(self, url, file_path):
-        response = requests.get(url)
-        with open(file_path, "wb") as f:
-            f.write(response.content)
+        with self.lock:  # ロックを取得
+            response = requests.get(url)
+            with open(file_path, "wb") as f:
+                f.write(response.content)
 
     def print_thread_info(self, start_time, end_time, duration):
         print(f"Thread {self.index} output_url: {self.output_url}")
@@ -165,7 +167,7 @@ class Video(Replicate):
 
 class Music(Replicate):
 
-    def __init__(self, id, client: Client, args):
+    def __init__(self, id, client: Client, args, duration):
         super().__init__(id, client, args)
         self.REPLICATE_MODEL_PATH = "facebookresearch/musicgen"
         self.REPLICATE_MODEL_VERSION = "f8578df960c345df7bc1f85dd152c5ae0b57ce45a6fc09511c467a62ad820ba3",
@@ -174,7 +176,7 @@ class Music(Replicate):
         
         self.file_path_format = "assets/{id}/{class_name}_{index}_request_{prediction_id}.mp3"
         self.file_path = None
-        self.duration = args.get("")
+        self.duration = duration
         self.input = {
             "model_version": "large", 
             "prompt": self.prompt,
@@ -229,9 +231,10 @@ class ThreadController:
         for token_index, token in enumerate(REPLICATE_API_TOKEN_LIST):
             client = Client()
             client.api_token = token
+            client.api_token_index = 0
             self.replicate_client_list[token] = client
             if token_index == 0:
-                self.music = Music(self.id, client, args)
+                self.music = Music(self.id, client, args, self.duration)
                 self.music.duration = self.duration
 
             for index, scene in enumerate(scenes):
@@ -239,7 +242,7 @@ class ThreadController:
                 video = Video(self.id, client, args, scene, index)
                 self.videos.append(video)
             
-            # client.api_token_index = (token_index + 1) % len(REPLICATE_API_TOKEN_LIST)
+            client.api_token_index = (token_index + 1) % len(REPLICATE_API_TOKEN_LIST)
 
     def run_threads(self):
 
@@ -265,7 +268,7 @@ class ThreadController:
 
     def merge_videos(self):
         clips = []
-        for video in self.videos:
+        for video in sorted(self.videos, key=lambda x: x.index):
             video_path = Path(video.file_path)
             if video_path.exists():
                 clips.append(VideoFileClip(video.file_path))
@@ -294,10 +297,12 @@ class ThreadController:
         # Trimming the final audio segment to match the video duration exactly
         final_audio_segment = final_audio_segment[:int(video_duration * 1000)]
 
-        temp_audio_path = "/tmp/temp_audio.mp3"
+        import tempfile
 
         # Saving the final audio as a temporary WAV file
-        final_audio_segment.export(temp_audio_path, format="mp3")
+        with tempfile.NamedTemporaryFile(suffix='.mp3', delete=False) as f:
+            temp_audio_path = f.name
+            final_audio_segment.export(temp_audio_path, format="mp3")
 
         # Loading the temporary audio file as a MoviePy AudioFileClip
         final_audio_clip = AudioFileClip(temp_audio_path)
@@ -305,14 +310,11 @@ class ThreadController:
         # Setting the audio to the video
         final_video_clip = video_clip.set_audio(final_audio_clip)
 
-        # Path to save the final video with audio (different name to avoid confusion)
-        output_path_with_audio_fixed = "/tmp/final_video_with_audio_fixed.mp4"
-
-        # Saving the final video with audio
-        final_video_clip.write_videofile(output_path_with_audio_fixed, codec="libx264", audio_codec="aac")
+        # Saving the final video with audio to a temporary file
+        with tempfile.NamedTemporaryFile(suffix='.mp4', delete=False) as f:
+            output_path_with_audio_fixed = f.name
+            final_video_clip.write_videofile(output_path_with_audio_fixed, codec="libx264", audio_codec="aac")
 
-        # Path to the final video with audio (fixed version)
-        output_path_with_audio_fixed
 
         os.makedirs(f"videos/{self.id}/", exist_ok=True)
         output_path = f"videos/{self.id}/final_concatenated_video_{self.id}.mp4"
@@ -321,6 +323,10 @@ class ThreadController:
         
         import shutil
         shutil.move(output_path_with_audio_fixed, output_path)
+        
+        # Delete the temporary files
+        os.remove(temp_audio_path)
+        shutil.rmtree(f"assets/{self.id}/")
 
         return output_path
 
@@ -468,6 +474,11 @@ if __name__ == "__main__":
         NajiminoAI.generate("伝統工芸と最新技術の融合")
     else:
 
+        description = """
+            入力されたテキストプロンプトに基づ��てビデオを生成します。
+            Generate a video based on the text prompt you enter.
+            """
+
         iface = gr.Interface(
             fn=NajiminoAI.generate,
             # inputs=gr.Textbox(label=inputs_label),
@@ -477,8 +488,8 @@ if __name__ == "__main__":
                 ],
             # title=title,
             inputs=gr.inputs.Textbox(lines=2, placeholder="Enter your prompt"),
-            title="Video Generator",
-            description="Generate a video based on the text prompt you enter.",
+            title="najimino Video Generator (β)",
+            description=description,
             examples=[
                 ["侍たちは野を超え山を超え、敵軍大将を討ち取り、天下の大将軍となった！"],
                 ["子どもたちが笑ったり怒ったり泣いたり楽しんだりする"],
@@ -487,13 +498,3 @@ if __name__ == "__main__":
         )
         iface.launch()
 
-# import replicate
-# import os
-# token = os.environ.get("REPLICATE_API_TOKEN")
-# os.environ["REPLICATE_API_TOKEN"] = token
-# print(f"token: {token}")
-# output = replicate.run(
-#     "facebookresearch/musicgen:7a76a8258b23fae65c5a22debb8841d1d7e816b75c2f24218cd2bd8573787906",
-#     input={"model_version": "melody"}
-# )
-# print(output)