import gradio as gr from openai import OpenAI from pydub import AudioSegment import numpy as np def transcribe_audio(audio_file, api_key): client = OpenAI(api_key=api_key) audio = AudioSegment.from_mp3(audio_file) segment_duration = 10 * 60 * 1000 num_segments = (len(audio) // segment_duration) + 1 segments = [] for i in range(num_segments): start_time = i * segment_duration end_time = (i + 1) * segment_duration segment = audio[start_time:end_time] segments.append(segment) print(f"Segment {i}: {start_time} - {end_time} ({len(segment)} ms)") text = "" for i, segment in enumerate(segments): segment.export('./segment.mp3', format="mp3") audio = open("./segment.mp3", 'rb') transcription = client.audio.transcriptions.create(file=audio, language='es', model="whisper-1") text += transcription.text + " " return text def main(): audio_input = gr.Audio(sources="upload", type="filepath") api_key = gr.Textbox(lines=1, label="OpenAI API Key") output_text = gr.Textbox() iface = gr.Interface(fn=transcribe_audio, inputs=[audio_input, api_key], outputs=output_text, title="Audio Transcription App", description="Upload an audio file and hit the 'Submit'\ button") iface.launch() if __name__ == '__main__': main()