app.py

import gradio as gr
from transformers import pipeline
import torch
import librosa
import os

# Authentication token for Hugging Face
auth_token = os.environ.get("HF_TOKEN")

# Mapping of language options to their codes
target_lang_options = {"English": "eng", "Luganda": "lug", "Acholi": "ach", "Runyankole": "nyn", "Lugbara": "lgg"}
languages = list(target_lang_options.keys())

# Determine device based on CUDA availability
device = "cuda" if torch.cuda.is_available() else "cpu"

base_model_id = "facebook/mms-1b-all"
pipe = pipeline("automatic-speech-recognition", model=base_model_id, device=device, token=auth_token)

def transcribe_audio(input_file, language, chunk_length_s=10, stride_length_s=(4, 2), return_timestamps="word"):
    target_lang_code = target_lang_options[language]

    # Dynamically set the target language and load the corresponding adapter
    pipe.tokenizer.set_target_lang(target_lang_code)
    # Assuming each language code directly corresponds to an adapter name available for the model
    pipe.model.load_adapter(target_lang_code)

    # Load and transcribe the audio file
    audio_data, _ = librosa.load(input_file, sr=None)
    output = pipe(audio_data, chunk_length_s=chunk_length_s, stride_length_s=stride_length_s, return_timestamps=return_timestamps)
    return output

description = "ASR with dynamic language adaptation"

iface = gr.Interface(fn=transcribe_audio,
                     inputs=[
                         gr.Audio(source="upload", type="filepath", label="Upload file to transcribe"),
                         gr.Dropdown(choices=languages, label="Language", value="English")
                     ],
                     outputs=gr.Textbox(label="Transcription"),
                     description=description)

iface.launch()