app.py

import gradio as gr
import nemo.collections.asr as nemo_asr
from pydub import AudioSegment
import os
import yt_dlp as youtube_dl
from huggingface_hub import login
from hazm import Normalizer
import numpy as np
import re
import time

# Fetch the token from an environment variable
HF_TOKEN = os.getenv("HF_TOKEN")
if not HF_TOKEN:
    raise ValueError("HF_TOKEN environment variable not set. Please provide a valid Hugging Face token.")

# Authenticate with Hugging Face
login(HF_TOKEN)

# Load the private NeMo ASR model
try:
    asr_model = nemo_asr.models.EncDecHybridRNNTCTCBPEModel.from_pretrained(
        model_name="faimlab/stt_fa_fastconformer_hybrid_large_dataset_v30"
    )
except Exception as e:
    raise RuntimeError(f"Failed to load model: {str(e)}")

normalizer = Normalizer()

def load_audio(audio_path):
    audio = AudioSegment.from_file(audio_path)
    audio = audio.set_channels(1).set_frame_rate(16000)
    audio_samples = np.array(audio.get_array_of_samples(), dtype=np.float32)
    audio_samples /= np.max(np.abs(audio_samples))
    return audio_samples, audio.frame_rate

def transcribe_chunk(audio_chunk, model):
    transcription = model.transcribe([audio_chunk], batch_size=1, verbose=False)
    return transcription[0].text

def transcribe_audio(file_path, model, chunk_size=30*16000):
    waveform, _ = load_audio(file_path)
    transcriptions = []
    for start in range(0, len(waveform), chunk_size):
        end = min(len(waveform), start + chunk_size)
        transcription = transcribe_chunk(waveform[start:end], model)
        transcriptions.append(transcription)

    transcriptions = ' '.join(transcriptions)
    transcriptions = re.sub(' +', ' ', transcriptions)
    transcriptions = normalizer.normalize(transcriptions)
    
    return transcriptions

# YouTube audio download function
YT_LENGTH_LIMIT_S = 3600  

def download_yt_audio(yt_url, filename, cookie_file="cookies.txt"):
    info_loader = youtube_dl.YoutubeDL()
    
    try:
        info = info_loader.extract_info(yt_url, download=False)
    except youtube_dl.utils.DownloadError as err:
        raise gr.Error(str(err))
    
    file_length = info["duration_string"]
    file_h_m_s = file_length.split(":")
    file_h_m_s = [int(sub_length) for sub_length in file_h_m_s]
    
    if len(file_h_m_s) == 1:
        file_h_m_s.insert(0, 0)
    if len(file_h_m_s) == 2:
        file_h_m_s.insert(0, 0)
    file_length_s = file_h_m_s[0] * 3600 + file_h_m_s[1] * 60 + file_h_m_s[2]
    
    if file_length_s > YT_LENGTH_LIMIT_S:
        yt_length_limit_hms = time.strftime("%HH:%MM:%SS", time.gmtime(YT_LENGTH_LIMIT_S))
        file_length_hms = time.strftime("%HH:%MM:%SS", time.gmtime(file_length_s))
        raise gr.Error(f"Maximum YouTube length is {yt_length_limit_hms}, got {file_length_hms} YouTube video.")
    
    ydl_opts = {"outtmpl": filename, "format": "worstvideo[ext=mp4]+bestaudio[ext=m4a]/best[ext=mp4]/best", "cookies": cookie_file}
    
    with youtube_dl.YoutubeDL(ydl_opts) as ydl:
        try:
            ydl.download([yt_url])
        except youtube_dl.utils.ExtractorError as err:
            raise gr.Error(str(err))


# Gradio Interface
def transcribe(audio):
    if audio is None:
        return "Please upload an audio file."
    
    transcription = transcribe_audio(audio, asr_model)

    return transcription

def transcribe_yt(yt_url):
    temp_filename = "/tmp/yt_audio.mp4"  # Temporary filename for the downloaded video
    download_yt_audio(yt_url, temp_filename)
    transcription = transcribe_audio(temp_filename, asr_model)
    return transcription

mf_transcribe = gr.Interface(
    fn=transcribe,
    inputs=gr.Microphone(type="filepath"),
    outputs=gr.Textbox(label="Transcription"),
    theme="huggingface",
    title="Persian ASR Transcription with NeMo Fast Conformer",
    description=(
        "Transcribe long-form microphone or audio inputs with the click of a button! Demo uses the NeMo's Fast Conformer Hybrid Large.\n\n"
        "Trained on ~800 hours of Persian speech dataset (Common Voice 17 (~300 hours), YouTube (~400 hours), NasleMana (~90 hours), In-house dataset (~70 hours)).\n\n"
        "For commercial applications, contact us via email: <[email protected]>.\n\n"
        "Credit FAIM Group, Sharif University of Technology.\n\n"
    ),
    allow_flagging="never",
)

# File upload tab
file_transcribe = gr.Interface(
    fn=transcribe,
    inputs=gr.Audio(type="filepath", label="Audio file"),
    outputs=gr.Textbox(label="Transcription"),
    theme="huggingface",
    title="Persian ASR Transcription with NeMo Fast Conformer",
    description=(
        "Transcribe long-form microphone or audio inputs with the click of a button! Demo uses the NeMo's Fast Conformer Hybrid Large.\n\n"
        "Trained on ~800 hours of Persian speech dataset (Common Voice 17 (~300 hours), YouTube (~400 hours), NasleMana (~90 hours), In-house dataset (~70 hours)).\n\n"
        "For commercial applications, contact us via email: <[email protected]>.\n\n"
        "Credit FAIM Group, Sharif University of Technology.\n\n"
    ),
    allow_flagging="never",
)

# YouTube tab
yt_transcribe = gr.Interface(
    fn=transcribe_yt,
    inputs=gr.Textbox(label="YouTube URL", placeholder="Enter the YouTube URL here"),
    outputs=gr.Textbox(label="Transcription"),
    theme="huggingface",
    title="Transcribe YouTube Video",
    description="Transcribe audio from a YouTube video by providing its URL. Currently YouTube is blocking the requests. So you will see the app showing error",
    allow_flagging="never",
)

# Gradio Interface
demo = gr.Blocks()

with demo:
    # Create the tabs with the list of interfaces
    gr.TabbedInterface([mf_transcribe, file_transcribe, yt_transcribe], ["Microphone", "Audio file", "YouTube"])

demo.launch()