app.py

import os
import time
import requests
import gradio as gr
from huggingface_hub import login
from dotenv import load_dotenv

load_dotenv()
image_API="https://api-inference.huggingface.co/models/stabilityai/stable-diffusion-2"
translation_API="https://api-inference.huggingface.co/models/facebook/nllb-200-distilled-1.3B"
whisper_API="https://api-inference.huggingface.co/models/openai/whisper-large-v3"

HF_TOKEN=os.getenv("HF_TOKEN")
login(HF_TOKEN)
headers={"Authorization":f"Bearer {HF_TOKEN}"}


def query_whisper(audio_path):
  with open(audio_path,"rb") as audio:
    data=audio.read()
    response=requests.post(whisper_API,headers=headers,data=data)
    return response.json()

def query_translation(text):
  max_retries=5
  delay=10
  payload={"inputs":text,"parameters":{"src_lang":"ta_Taml","tgt_lang":'eng_Latn'}}
  for attempt in range(max_retries):
    response=requests.post(translation_API,headers=headers,json=payload)
    result= response.json()
    if isinstance(result,list) and len(result)>0:
      return {"translated_text":result[0]["translation_text"]}
    elif isinstance(result,dict) and "translation_text" in result:
      return {"translated_text":result["translation_text"]}
    time.sleep(delay)
  return {"error":"Translation API failed","response":result}

def query_image(prompt):
  max_retries=5
  delay=10
  payload={"inputs":prompt}
  for attempt in range(max_retries):
    response=requests.post(image_API,headers=headers,json=payload)
    if response.status_code == 200 and response.headers.get('Content-Type', '').startswith('image/'):
      image_path='generated_image.png'
      with open(image_path,'wb') as f:
        f.write(response.content)
      return image_path
    time.sleep(delay)
  return None


def process_audio(audio_path):
  if not audio_path:
    return None
  try:
    transcription=query_whisper(audio_path)
    tamil_text=transcription.get("text","Transcription error")
    translation=query_translation(tamil_text)
    translated_text=translation.get("translated_text","Translation error")
    image_path=query_image(translated_text)
    return tamil_text,translated_text,image_path
  except Exception as e:
    return None,str(e),None

iface=gr.Interface(
    fn=process_audio,
    inputs=gr.Audio(type="filepath",label="Upload Audio"),
    outputs=[
        gr.Textbox(label="Tamil Text"),
        gr.Textbox(label="English Translation"),
        gr.Image(label="Generated Image")
        
    ],
    title="Speech-to-Image Generation"
)
iface.launch(share=True,server_name="0.0.0.0",server_port=None)