Duplicate from nanomenta/Whisper_speaker_diarization

.gitattributes

@@ -0,0 +1,34 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,15 @@
+---
+title: Whisper Speaker Diarization
+emoji: 🎎
+colorFrom: blue
+colorTo: red
+sdk: gradio
+sdk_version: 3.9.1
+app_file: app.py
+pinned: false
+tags:
+- whisper-event
+duplicated_from: nanomenta/Whisper_speaker_diarization
+---
+
+Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

app.py

@@ -0,0 +1,338 @@
+import whisper
+import datetime
+import subprocess
+import gradio as gr
+from pathlib import Path
+import pandas as pd
+import re
+import time
+import os 
+import numpy as np
+from sklearn.cluster import AgglomerativeClustering
+from sklearn.metrics import silhouette_score
+
+from pytube import YouTube
+import torch
+import pyannote.audio
+from pyannote.audio.pipelines.speaker_verification import PretrainedSpeakerEmbedding
+from pyannote.audio import Audio
+from pyannote.core import Segment
+
+from gpuinfo import GPUInfo
+
+import wave
+import contextlib
+from transformers import pipeline
+import psutil
+
+whisper_models = ["base", "small", "medium", "large"]
+source_languages = {
+    "en": "English",
+    "zh": "Chinese",
+    "de": "German",
+    "es": "Spanish",
+    "ru": "Russian",
+    "ko": "Korean",
+    "fr": "French",
+    "ja": "Japanese",
+    "pt": "Portuguese",
+    "tr": "Turkish",
+    "pl": "Polish",
+    "ca": "Catalan",
+    "nl": "Dutch",
+    "ar": "Arabic",
+    "sv": "Swedish",
+    "it": "Italian",
+    "id": "Indonesian",
+    "hi": "Hindi",
+    "fi": "Finnish",
+    "vi": "Vietnamese",
+    "he": "Hebrew",
+    "uk": "Ukrainian",
+    "el": "Greek",
+    "ms": "Malay",
+    "cs": "Czech",
+    "ro": "Romanian",
+    "da": "Danish",
+    "hu": "Hungarian",
+    "ta": "Tamil",
+    "no": "Norwegian",
+    "th": "Thai",
+    "ur": "Urdu",
+    "hr": "Croatian",
+    "bg": "Bulgarian",
+    "lt": "Lithuanian",
+    "la": "Latin",
+    "mi": "Maori",
+    "ml": "Malayalam",
+    "cy": "Welsh",
+    "sk": "Slovak",
+    "te": "Telugu",
+    "fa": "Persian",
+    "lv": "Latvian",
+    "bn": "Bengali",
+    "sr": "Serbian",
+    "az": "Azerbaijani",
+    "sl": "Slovenian",
+    "kn": "Kannada",
+    "et": "Estonian",
+    "mk": "Macedonian",
+    "br": "Breton",
+    "eu": "Basque",
+    "is": "Icelandic",
+    "hy": "Armenian",
+    "ne": "Nepali",
+    "mn": "Mongolian",
+    "bs": "Bosnian",
+    "kk": "Kazakh",
+    "sq": "Albanian",
+    "sw": "Swahili",
+    "gl": "Galician",
+    "mr": "Marathi",
+    "pa": "Punjabi",
+    "si": "Sinhala",
+    "km": "Khmer",
+    "sn": "Shona",
+    "yo": "Yoruba",
+    "so": "Somali",
+    "af": "Afrikaans",
+    "oc": "Occitan",
+    "ka": "Georgian",
+    "be": "Belarusian",
+    "tg": "Tajik",
+    "sd": "Sindhi",
+    "gu": "Gujarati",
+    "am": "Amharic",
+    "yi": "Yiddish",
+    "lo": "Lao",
+    "uz": "Uzbek",
+    "fo": "Faroese",
+    "ht": "Haitian creole",
+    "ps": "Pashto",
+    "tk": "Turkmen",
+    "nn": "Nynorsk",
+    "mt": "Maltese",
+    "sa": "Sanskrit",
+    "lb": "Luxembourgish",
+    "my": "Myanmar",
+    "bo": "Tibetan",
+    "tl": "Tagalog",
+    "mg": "Malagasy",
+    "as": "Assamese",
+    "tt": "Tatar",
+    "haw": "Hawaiian",
+    "ln": "Lingala",
+    "ha": "Hausa",
+    "ba": "Bashkir",
+    "jw": "Javanese",
+    "su": "Sundanese",
+}
+
+source_language_list = [key[0] for key in source_languages.items()]
+
+os.makedirs('output', exist_ok=True)
+
+embedding_model = PretrainedSpeakerEmbedding( 
+    "speechbrain/spkrec-ecapa-voxceleb",
+    device=torch.device("cuda" if torch.cuda.is_available() else "cpu"))
+
+
+
+def convert_time(secs):
+    return datetime.timedelta(seconds=round(secs))
+
+
+
+def speech_to_text(audio_file_path, selected_source_lang, whisper_model, num_speakers):
+    """
+    # Transcribe youtube link using OpenAI Whisper
+    1. Using Open AI's Whisper model to seperate audio into segments and generate transcripts.
+    2. Generating speaker embeddings for each segments.
+    3. Applying agglomerative clustering on the embeddings to identify the speaker for each segment.
+    
+    Speech Recognition is based on models from OpenAI Whisper https://github.com/openai/whisper
+    Speaker diarization model and pipeline from by https://github.com/pyannote/pyannote-audio
+    """
+    
+    model = whisper.load_model(whisper_model)
+    time_start = time.time()
+    if(audio_file_path == None):
+        raise ValueError("Error no audio input")
+    print(audio_file_path)
+
+    try:
+        # Read and convert youtube video
+        #_,file_ending = os.path.splitext(f'{video_file_path}')
+        #print(f'file enging is {file_ending}')
+        #audio_file = video_file_path.replace(file_ending, ".wav")
+        #print("starting conversion to wav")
+        #os.system(f'ffmpeg -i "{video_file_path}" -ar 16000 -ac 1 -c:a pcm_s16le "{audio_file}"')
+
+        audio_file = audio_file_path
+        
+        # Get duration
+        with contextlib.closing(wave.open(audio_file,'r')) as f:
+            frames = f.getnframes()
+            rate = f.getframerate()
+            duration = frames / float(rate)
+        print(f"conversion to wav ready, duration of audio file: {duration}")
+
+        # Transcribe audio
+        options = dict(language=selected_source_lang, beam_size=5, best_of=5)
+        transcribe_options = dict(task="transcribe", **options)
+        result = model.transcribe(audio_file, **transcribe_options)
+        segments = result["segments"]
+        print("starting whisper done with whisper")
+    except Exception as e:
+        raise RuntimeError("Error converting video to audio")
+
+    try:
+        # Create embedding
+        def segment_embedding(segment):
+            audio = Audio()
+            start = segment["start"]
+            # Whisper overshoots the end timestamp in the last segment
+            end = min(duration, segment["end"])
+            clip = Segment(start, end)
+            waveform, sample_rate = audio.crop(audio_file, clip)
+            return embedding_model(waveform[None])
+
+        embeddings = np.zeros(shape=(len(segments), 192))
+        for i, segment in enumerate(segments):
+            embeddings[i] = segment_embedding(segment)
+        embeddings = np.nan_to_num(embeddings)
+        print(f'Embedding shape: {embeddings.shape}')
+
+        if num_speakers == 0:
+        # Find the best number of speakers
+            score_num_speakers = {}
+    
+            for num_speakers in range(2, 10+1):
+                clustering = AgglomerativeClustering(num_speakers).fit(embeddings)
+                score = silhouette_score(embeddings, clustering.labels_, metric='euclidean')
+                score_num_speakers[num_speakers] = score
+            best_num_speaker = max(score_num_speakers, key=lambda x:score_num_speakers[x])
+            print(f"The best number of speakers: {best_num_speaker} with {score_num_speakers[best_num_speaker]} score")
+        else:
+            best_num_speaker = num_speakers
+            
+        # Assign speaker label   
+        clustering = AgglomerativeClustering(best_num_speaker).fit(embeddings)
+        labels = clustering.labels_
+        for i in range(len(segments)):
+            segments[i]["speaker"] = 'SPEAKER ' + str(labels[i] + 1)
+
+        # Make output
+        objects = {
+            'Start' : [],
+            'End': [],
+            'Speaker': [],
+            'Text': []
+        }
+        text = ''
+        for (i, segment) in enumerate(segments):
+            if i == 0 or segments[i - 1]["speaker"] != segment["speaker"]:
+                objects['Start'].append(str(convert_time(segment["start"])))
+                objects['Speaker'].append(segment["speaker"])
+                if i != 0:
+                    objects['End'].append(str(convert_time(segments[i - 1]["end"])))
+                    objects['Text'].append(text)
+                    text = ''
+            text += segment["text"] + ' '
+        objects['End'].append(str(convert_time(segments[i - 1]["end"])))
+        objects['Text'].append(text)
+        
+        time_end = time.time()
+        time_diff = time_end - time_start
+        memory = psutil.virtual_memory()
+        gpu_utilization, gpu_memory = GPUInfo.gpu_usage()
+        gpu_utilization = gpu_utilization[0] if len(gpu_utilization) > 0 else 0
+        gpu_memory = gpu_memory[0] if len(gpu_memory) > 0 else 0
+        system_info = f"""
+        *Memory: {memory.total / (1024 * 1024 * 1024):.2f}GB, used: {memory.percent}%, available: {memory.available / (1024 * 1024 * 1024):.2f}GB.* 
+        *Processing time: {time_diff:.5} seconds.*
+        *GPU Utilization: {gpu_utilization}%, GPU Memory: {gpu_memory}MiB.*
+        """
+        save_path = "output/transcript_result.csv"
+        df_results = pd.DataFrame(objects)
+        df_results.to_csv(save_path)
+        return df_results, system_info, save_path
+    
+    except Exception as e:
+        raise RuntimeError("Error Running inference with local model", e)
+
+
+# ---- Gradio Layout -----
+# Inspiration from https://huggingface.co/spaces/RASMUS/Whisper-youtube-crosslingual-subtitles
+audio_in = gr.Audio(label="Audio file", type="filepath", source="upload")
+
+df_init = pd.DataFrame(columns=['Start', 'End', 'Speaker', 'Text'])
+memory = psutil.virtual_memory()
+selected_source_lang = gr.Dropdown(choices=source_language_list, type="value", value="en", label="Spoken language in audio", interactive=True)
+selected_whisper_model = gr.Dropdown(choices=whisper_models, type="value", value="base", label="Selected Whisper model", interactive=True)
+number_speakers = gr.Number(precision=0, value=0, label="Input number of speakers for better results. If value=0, model will automatic find the best number of speakers", interactive=True)
+system_info = gr.Markdown(f"*Memory: {memory.total / (1024 * 1024 * 1024):.2f}GB, used: {memory.percent}%, available: {memory.available / (1024 * 1024 * 1024):.2f}GB*")
+download_transcript = gr.File(label="Download transcript")
+transcription_df = gr.DataFrame(value=df_init,label="Transcription dataframe", row_count=(0, "dynamic"), max_rows = 10, wrap=True, overflow_row_behaviour='paginate')
+title = "Whisper speaker diarization"
+demo = gr.Blocks(title=title)
+demo.encrypt = False
+
+
+with demo:
+    with gr.Tab("Whisper speaker diarization"):
+        gr.Markdown('''
+            <div>
+            <h1 style='text-align: center'>Whisper speaker diarization</h1>
+            This space uses Whisper models from <a href='https://github.com/openai/whisper' target='_blank'><b>OpenAI</b></a> to recoginze the speech and ECAPA-TDNN model from <a href='https://github.com/speechbrain/speechbrain' target='_blank'><b>SpeechBrain</b></a> to encode and clasify speakers</h2>
+            </div>
+        ''')
+
+        with gr.Row():
+            gr.Markdown('''
+            ### Transcribe audio files using OpenAI Whisper
+            ##### 1. Using Open AI's Whisper model to seperate audio into segments and generate transcripts.
+            ##### 2. Generating speaker embeddings for each segments.
+            ##### 3. Applying agglomerative clustering on the embeddings to identify the speaker for each segment.
+            ''')
+              
+
+        
+                
+
+        with gr.Row():
+            with gr.Column():
+                audio_in.render()
+                with gr.Column():
+                    gr.Markdown('''
+                    ##### Here you can start the transcription process.
+                    ##### Please select the source language for transcription.
+                    ##### You can select a range of assumed numbers of speakers.
+                    ''')
+                selected_source_lang.render()
+                selected_whisper_model.render()
+                number_speakers.render()
+                transcribe_btn = gr.Button("Transcribe audio and diarization")
+                transcribe_btn.click(speech_to_text, 
+                                     [audio_in, selected_source_lang, selected_whisper_model, number_speakers], 
+                                     [transcription_df, system_info, download_transcript]
+                                    )
+                
+        with gr.Row():
+            gr.Markdown('''
+            ##### Here you will get transcription  output
+            ##### ''')
+            
+
+        with gr.Row():
+            with gr.Column():
+                download_transcript.render()
+                transcription_df.render()
+                system_info.render()
+                gr.Markdown('''<center><img src='https://visitor-badge.glitch.me/badge?page_id=WhisperDiarizationSpeakers' alt='visitor badge'><a href="https://opensource.org/licenses/Apache-2.0"><img src='https://img.shields.io/badge/License-Apache_2.0-blue.svg' alt='License: Apache 2.0'></center>''')
+
+                
+    
+
+
+demo.launch(debug=True)

packages.txt

@@ -0,0 +1 @@
+ffmpeg

requirements.txt

@@ -0,0 +1,20 @@
+git+https://github.com/huggingface/transformers
+git+https://github.com/pyannote/pyannote-audio
+git+https://github.com/openai/whisper.git
+gradio==3.12
+ffmpeg-python
+pandas==1.5.0
+pytube==12.1.0
+sacremoses
+sentencepiece
+tokenizers
+torch
+torchaudio
+tqdm==4.64.1
+EasyNMT==2.0.2   
+nltk
+transformers
+pysrt
+psutil==5.9.2
+requests
+gpuinfo