app.py

import gradio as gr
import requests
import numpy as np
from pydub import AudioSegment
import io
from IPython.display import Audio

# Define the Hugging Face Inference API URLs and headers
ASR_API_URL = "https://api-inference.huggingface.co/models/Baghdad99/saad-speech-recognition-hausa-audio-to-text"
TTS_API_URL = "https://api-inference.huggingface.co/models/Baghdad99/english_voice_tts"
TRANSLATION_API_URL = "https://api-inference.huggingface.co/models/Baghdad99/saad-hausa-text-to-english-text"
headers = {"Authorization": "Bearer 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"}

# Define the function to query the Hugging Face Inference API
def query(api_url, payload=None, data=None):
    if data is not None:
        response = requests.post(api_url, headers=headers, data=data)
    else:
        response = requests.post(api_url, headers=headers, json=payload)
    response_json = response.json()
    if 'error' in response_json:
        print(f"Error in query function: {response_json['error']}")
        return None
    return response_json


# Define the function to translate speech
def translate_speech(audio_file):
    print(f"Type of audio: {type(audio_file)}, Value of audio: {audio_file}")  # Debug line

    # Use the ASR pipeline to transcribe the audio
    data = audio_file.read()
    output = query(ASR_API_URL, data=data)
    print(f"Output: {output}")  # Debug line

    # Check if 'error' key exists in the output
    if 'error' in output:
        print(f"Error: {output['error']}")
        estimated_time = output.get('estimated_time')
        if estimated_time:
            print(f"Estimated time for the model to load: {estimated_time} seconds")
        return

    # Check if 'text' key exists in the output
    if 'text' in output:
        transcription = output["text"]
    else:
        print("Key 'text' does not exist in the output.")
        return

    # Use the translation pipeline to translate the transcription
    translated_text = query(TRANSLATION_API_URL, {"inputs": transcription})

    # Use the TTS pipeline to synthesize the translated text
    response = requests.post(TTS_API_URL, headers=headers, json={"inputs": translated_text})
    audio_bytes = response.content

    # Display the audio output
    return Audio(audio_bytes)

    # print(f"Type of audio: {type(audio_file)}, Value of audio: {audio_file}")  # Debug line

    # # Use the ASR pipeline to transcribe the audio
    # data = audio_file.read()
    # output = query(ASR_API_URL, data=data)
    # print(f"Output: {output}")  # Debug line

    # # Check if 'text' key exists in the output
    # if 'text' in output:
    #     transcription = output["text"]
    # else:
    #     print("Key 'text' does not exist in the output.")
    #     return

    # # Use the translation pipeline to translate the transcription
    # translated_text = query(TRANSLATION_API_URL, {"inputs": transcription})

    # # Use the TTS pipeline to synthesize the translated text
    # response = requests.post(TTS_API_URL, headers=headers, json={"inputs": translated_text})
    # audio_bytes = response.content

    # # Display the audio output
    # return Audio(audio_bytes)


# Define the Gradio interface
iface = gr.Interface(
    fn=translate_speech, 
    inputs=gr.inputs.File(type="file"),  # Change this line
    outputs=gr.outputs.Audio(type="numpy"),
    title="Hausa to English Translation",
    description="Realtime demo for Hausa to English translation using speech recognition and text-to-speech synthesis."
)

iface.launch()