Update app.py

app.py

@@ -1,41 +1,11 @@
-import subprocess
-import ffmpeg
-import numpy as np
+from transformers import GPTNeoForCausalLM, GPT2Tokenizer
 
-# Upload the video file
-video_file = input("Enter the path to the video file: ")
+model = GPTNeoForCausalLM.from_pretrained("EleutherAI/gpt-neo-2.7B")
+tokenizer = GPT2Tokenizer.from_pretrained("EleutherAI/gpt-neo-2.7B")
 
-# Trim the video to 5 seconds
-ffmpeg.input(video_file).trim(start=0, duration=5).output("preview2.mp4").run()
+prompt = "The quick brown fox"
+input_ids = tokenizer(prompt, return_tensors="pt").input_ids
+generated_text = model.generate(input_ids=input_ids, max_length=50, do_sample=True)
 
-# Load the audio from the trimmed video
-audio, _ = (ffmpeg
-            .input("preview2.mp4")
-            .filter("aextract")
-            .output("-", format="s16le", acodec="pcm_s16le")
-            .run(capture_stdout=True, quiet=True)
-          )
-audio = np.frombuffer(audio, dtype=np.int16)
-
-# Define the pitch progression
-pitch_progression = [0, 0, 1, 1, -2, -2, 1, 1]
-
-# Create the final audio by applying the pitch progression to the video audio
-final_audio = np.empty(0, dtype=np.int16)
-for i in range(len(pitch_progression)):
-    # Set the pitch shift
-    pitch_shift = pitch_progression[i]
-
-    # Apply the pitch shift to the audio
-    audio_shifted = np.roll(audio, pitch_shift)
-    
-    # Add the shifted audio to the final audio
-    final_audio = np.concatenate([final_audio, audio_shifted])
-
-# Export the final audio as an mp3 file
-ffmpeg.input('pipe:', format='s16le', acodec='pcm_s16le', ar='48000', channels=1, seekable='0', bufsize='512k', stdin_final=True).output("preview2.mp3", format="mp3").run(input=final_audio.tobytes())
-
-# Delete the temporary video file
-subprocess.run(['rm', 'preview2.mp4'])
-
-print("Preview 2 generated successfully!")
+decoded_text = tokenizer.decode(generated_text[0], skip_special_tokens=True)
+print(decoded_text)