solve conflict

Former-commit-id: e28f8408800aef82cd6bb3e5d4e980e96f286297

README.md

@@ -8,11 +8,11 @@ pip install -r requirement.txt
 
 ## Usage
 ```
-usage: pipeline.py [-h] [--link LINK] [--local_path LOCAL_PATH] [--download DOWNLOAD] [--result RESULT] [--video_name VIDEO_NAME]
+usage: pipeline.py [-h] [--link LINK] [--local_path LOCAL_PATH] [--download DOWNLOAD] [--result RESULT] [--video_name VIDEO_NAME] [--model_name]
 
 quick start:
 
-example online: python3 pipeline.py --link https://www.youtube.com/watch?v=XbgFIkhMM3s --download ./downloads --result ./results --video_name uncle_roger_test
+example online: python3 pipeline.py --link https://www.youtube.com/watch?v=61c4dn6851g --download ./downloads --result ./results --video_name SO_I_CHOSE_RANDOM
 
 example offline: python3 pipeline.py --local_path test_translation.m4a --result ./results --video_name test_translation
 
@@ -27,4 +27,10 @@ options:
   --result RESULT       translate result path
   --video_name VIDEO_NAME
                         video name
+  --model_name MODEL_NAME 
+                        model name
+
+if you cannot download youtube video, please follow the link below.
+https://github.com/pytube/pytube/issues/1498
+
 ```

pipeline.py

@@ -21,24 +21,27 @@ if args.link is None and args.local_path is None and args.srt_file is None:
     print("need video source or srt file")
     exit()
 
+# set openai api key
 openai.api_key = os.getenv("OPENAI_API_KEY")
 DOWNLOAD_PATH = args.download
 RESULT_PATH = args.output_dir
 VIDEO_NAME = args.video_name
-n_threshold = 1000  # Token limit for the GPT-3.5 model
-# model_name = "text-davinci-003" # replace this to our own fintune model
 model_name = args.model_name
-# model_name = "davinci"
 
 # get source audio
 if args.link is not None and args.local_path is None:
     # Download audio from YouTube
     video_link = args.link
+    video = None
+    audio = None
     try:
         video = YouTube(video_link)
         audio = video.streams.filter(only_audio=True, file_extension='mp4').first()
-        audio.download(DOWNLOAD_PATH)
-        print('Download Completed!')
+        if audio:
+            audio.download(DOWNLOAD_PATH)
+            print('Download Completed!')
+        else:
+            print("Error: Audio stream not found")
     except Exception as e:
         print("Connection Error")
         print(e) 
@@ -84,7 +87,7 @@ assSub_en = srt2ass(srt_file_en, "default", "No", "Modest")
 print('ASS subtitle saved as: ' + assSub_en)
 
 # Split the video script by sentences and create chunks within the token limit
-n_threshold = 4096  # Token limit for the GPT-3 model
+n_threshold = 1500  # Token limit for the GPT-3 model
 script_split = script_input.split('.')
 
 script_arr = []
@@ -109,7 +112,7 @@ for s in script_arr:
                 {"role": "system", "content": "You are a helpful assistant that translates English to Chinese and have decent background in starcraft2."},
                 {"role": "user", "content": 'Translate the following English text to Chinese: "{}"'.format(s)}
             ],
-            temperature=0.1
+            temperature=0.15
         )
         with open(f"{RESULT_PATH}/{VIDEO_NAME}_zh.srt", 'a+') as f:
             f.write(response['choices'][0]['message']['content'].strip())
@@ -131,4 +134,4 @@ for s in script_arr:
             f.write(response['choices'][0]['text'].strip())
 
 assSub_zh = srt2ass(f"{RESULT_PATH}/{VIDEO_NAME}_zh.srt", "default", "No", "Modest")
-print('ASS subtitle saved as: ' + assSub_zh)
+print('ASS subtitle saved as: ' + assSub_zh)