Tuesday, May 14, 2024

Use GPT-4o Locally for text, audio, video and transcription

 This video shows how to install and use GPT-4o API for audio video transcription and processing locally.




Code:

from openai import OpenAI
import os
import cv2
from moviepy.editor import VideoFileClip
import time
import base64

MODEL="gpt-4o"
os.environ.get('OPENAI_API_KEY')
client = OpenAI(api_key=os.environ.get('OPENAI_API_KEY'))

VIDEO_PATH = "myvideo.mp4"

def process_video(video_path, seconds_per_frame=2):
    base64Frames = []
    base_video_path, _ = os.path.splitext(video_path)

    video = cv2.VideoCapture(video_path)
    total_frames = int(video.get(cv2.CAP_PROP_FRAME_COUNT))
    fps = video.get(cv2.CAP_PROP_FPS)
    frames_to_skip = int(fps * seconds_per_frame)
    curr_frame=0

    # Loop through the video and extract frames at specified sampling rate
    while curr_frame < total_frames - 1:
        video.set(cv2.CAP_PROP_POS_FRAMES, curr_frame)
        success, frame = video.read()
        if not success:
            break
        _, buffer = cv2.imencode(".jpg", frame)
        base64Frames.append(base64.b64encode(buffer).decode("utf-8"))
        curr_frame += frames_to_skip
    video.release()

    # Extract audio from video
    audio_path = f"{base_video_path}.mp3"
    clip = VideoFileClip(video_path)
    clip.audio.write_audiofile(audio_path, bitrate="32k")
    clip.audio.close()
    clip.close()

    print(f"Extracted {len(base64Frames)} frames")
    print(f"Extracted audio to {audio_path}")
    return base64Frames, audio_path

# Extract 1 frame per second. You can adjust the `seconds_per_frame` parameter to change the sampling rate
base64Frames, audio_path = process_video(VIDEO_PATH, seconds_per_frame=1)

transcription = client.audio.transcriptions.create(
    model="whisper-1",
    file=open(audio_path, "rb"),
)

## Generate a summary with visual and audio
response = client.chat.completions.create(
    model=MODEL,
    messages=[
    {"role": "system", "content":"""You are generating a video summary. Create a summary of the provided video and its transcript. Respond in Markdown"""},
    {"role": "user", "content": [
        "These are the frames from the video.",
        *map(lambda x: {"type": "image_url",
                        "image_url": {"url": f'data:image/jpg;base64,{x}', "detail": "low"}}, base64Frames),
        {"type": "text", "text": f"The audio transcription is: {transcription.text}"}
        ],
    }
],
    temperature=0,
)
print(response.choices[0].message.content)

No comments: