import spaces import torch import gradio as gr import yt_dlp as youtube_dl from transformers import pipeline from transformers.pipelines.audio_utils import ffmpeg_read import tempfile import os MODEL_NAME = "TalTechNLP/whisper-large-v3-turbo-et-subs" BATCH_SIZE = 8 FILE_LIMIT_MB = 1000 YT_LENGTH_LIMIT_S = 3600 # limit to 1 hour YouTube files device = 0 if torch.cuda.is_available() else "cpu" pipe = pipeline( task="automatic-speech-recognition", model=MODEL_NAME, chunk_length_s=30, device=device, ) def convert_to_vtt(whisper_output): """ Convert Whisper ASR output to VTT subtitle format. Args: whisper_output (dict): Dictionary containing Whisper ASR output with 'text' and 'chunks' Returns: str: VTT formatted subtitles as a string """ def format_timestamp(seconds): """Convert seconds to VTT timestamp format (HH:MM:SS.mmm)""" if seconds is None: return "99:59:59.999" # Use max time for None values hours = int(seconds // 3600) minutes = int((seconds % 3600) // 60) seconds_remainder = seconds % 60 return f"{hours:02d}:{minutes:02d}:{seconds_remainder:06.3f}".replace('.', ',') # Start with VTT header vtt_output = "WEBVTT\n\n" # Process each chunk for i, chunk in enumerate(whisper_output['chunks'], 1): start_time, end_time = chunk['timestamp'] # Format the subtitle entry vtt_output += f"{i}\n" vtt_output += f"{format_timestamp(start_time)} --> {format_timestamp(end_time)}\n" vtt_output += f"{chunk['text'].strip()}\n\n" return vtt_output def dynamic_gpu_duration(func, duration, *args): @spaces.GPU(duration=duration) def wrapped_func(): return func(*args) return wrapped_func() @spaces.GPU def dummy_gpu(): return None def do_transcribe(inputs): if inputs is None: raise gr.Error("No audio file submitted! Please upload or record an audio file before submitting your request.") result = pipe(inputs, batch_size=BATCH_SIZE, generate_kwargs={"task": "transcribe", "language": "et"}, return_timestamps=True) return convert_to_vtt(result) def transcribe(file_path): with open(file_path, "rb") as f: audio_data = ffmpeg_read(f.read(), 16000) # Calculate the length in seconds audio_length = len(audio_data) / 16000 #expected_transcribe_duration = max(59, int(audio_length / 5.0)) expected_transcribe_duration = 59 gr.Info(f"Starting to transcribe, requesting a GPU for {expected_transcribe_duration} seconds") return dynamic_gpu_duration(do_transcribe, expected_transcribe_duration, file_path) def _return_yt_html_embed(yt_url): video_id = yt_url.split("?v=")[-1] HTML_str = ( f'
' "
" ) return HTML_str def download_yt_audio(yt_url, filename): 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"} 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)) def yt_transcribe(yt_url, max_filesize=75.0): with tempfile.TemporaryDirectory() as tmpdirname: filepath = os.path.join(tmpdirname, "video.mp4") download_yt_audio(yt_url, filepath) text = transcribe(transcribe, filepath) return text demo = gr.Blocks(theme=gr.themes.Ocean()) mf_transcribe = gr.Interface( fn=transcribe, inputs=[ gr.Audio(sources="microphone", type="filepath") ], #outputs="text", outputs=gr.Textbox(label="VTT subtitles", elem_id="text", show_label=True, show_copy_button=True, autoscroll=False, interactive=True), title="Generate Estonian subtitles", description=( "Transcribe long-form microphone or audio inputs with the click of a button! Demo uses the" f" checkpoint [{MODEL_NAME}](https://huggingface.co/{MODEL_NAME}) and 🤗 Transformers to transcribe audio files" " of arbitrary length." ), allow_flagging="never", ) file_transcribe = gr.Interface( fn=transcribe, inputs=[ gr.Audio(sources="upload", type="filepath", label="Audio file") ], #outputs="text", outputs=gr.Textbox(label="VTT subtitles", elem_id="text", show_label=True, show_copy_button=True, autoscroll=False, interactive=True), title="Generate Estonian subtitles", description=( "Transcribe long-form microphone or audio inputs with the click of a button! Demo uses the" f" checkpoint [{MODEL_NAME}](https://huggingface.co/{MODEL_NAME}) and 🤗 Transformers to transcribe audio files" " of arbitrary length." ), allow_flagging="never", ) yt_transcribe = gr.Interface( fn=yt_transcribe, inputs=[ gr.Textbox(lines=1, placeholder="Paste the URL to a YouTube video here", label="YouTube URL") ], #outputs=["html", "text"], outputs=gr.Textbox(label="VTT subtitles", elem_id="text", show_label=True, show_copy_button=True, autoscroll=False, interactive=True), title="Generate Estonian subtitles", description=( "Transcribe long-form YouTube videos with the click of a button! Demo uses the checkpoint" f" [{MODEL_NAME}](https://huggingface.co/{MODEL_NAME}) and 🤗 Transformers to transcribe video files of" " arbitrary length. NB! YouTube seems to often block download requests from Huggingface and there is nothing we can do about it." ), allow_flagging="never", ) with demo: gr.TabbedInterface([mf_transcribe, file_transcribe, yt_transcribe], ["Microphone", "Audio file", "YouTube"]) demo.queue().launch(ssr_mode=False)