app.py

import os
os.system("pip install git+https://github.com/openai/whisper.git")
import gradio as gr
import whisper
from flask import Flask, jsonify, request
import requests
import streamlit as st



model = whisper.load_model("small")

app = Flask(__name__)

@app.route("/")
def indexApi():
    return jsonify({"output": "okay"})

@app.route("/run", methods=['POST'])
def runApi():
    audio_url = request.form.get("audio_url")
    # key = request.form.get("key")
    # modelSelection = request.form.get("model")
    # print(audio_url)

    # if (modelSelection == None):
    #     modelSelection = "small"
    # model = whisper.load_model(modelSelection)
    # print(model)

    # # reject if key not the same 
    # apiKey = st.secrets["Api-Key"]
    # if apiKey != key:
    #     return jsonify({
    #         "image_url": image_url,
    #         "model": model,
    #         "result": "Invalid API Key",
    #     }), 400


    response = requests.get(audio_url)

    if response.status_code == requests.codes.ok:
        with open("audio.mp3", "wb") as f:
            f.write(response.content)
      
    else:
        return jsonify({
            "result": "Unable to save file, status code:  {response.status_code}" ,
        }), 400

    # arr = np.asarray(bytearray(response.content), dtype=np.uint8)
    # result = model.transcribe("audio.mp3")
    audio = "audio.mp3"

    audio = whisper.load_audio(audio)
    audio = whisper.pad_or_trim(audio)
    
    mel = whisper.log_mel_spectrogram(audio).to(model.device)
    
    _, probs = model.detect_language(mel)
    
    options = whisper.DecodingOptions(fp16 = False)
    result = whisper.decode(model, mel, options)
    print(result.text)
    return jsonify({
        "audio_url": audio_url,
        "model": model,
        # "result": result.text,
    })

if __name__ == "__main__":
    app.run(host="0.0.0.0", port=7860)
        
# def inference(audio):
#     audio = whisper.load_audio(audio)
#     audio = whisper.pad_or_trim(audio)
    
#     mel = whisper.log_mel_spectrogram(audio).to(model.device)
    
#     _, probs = model.detect_language(mel)
    
#     options = whisper.DecodingOptions(fp16 = False)
#     result = whisper.decode(model, mel, options)
    
#     # print(result.text)
#     return result.text, gr.update(visible=True), gr.update(visible=True), gr.update(visible=True)