Update app.py

app.py

@@ -1,6 +1,6 @@
-#!/usr/bin/env python
 ### -----------------------------------------------------------------------
-### (test_BASE, Revised) version_1.07 ALPHA, app.py
+### Transkriber version_1.00
+### app.py
 ### -----------------------------------------------------------------------
 
 # -------------------------------------------------------------------------
@@ -17,122 +17,135 @@
 # limitations under the License.
 # -------------------------------------------------------------------------
 
+
 import os
 import re
 import uuid
 import time
 import psutil
-import pydub
 import subprocess
 from tqdm import tqdm
-
 import tempfile
 from fpdf import FPDF
 from pathlib import Path
-
 import numpy as np
-import soundfile as sf
-import librosa
 import torch
-from transformers import pipeline 
-
+from transformers import pipeline
 from gpuinfo import GPUInfo
-
+from pydub import AudioSegment
+from IPython.display import Audio
 import gradio as gr
+import huggingface_hub
 
 
 ###############################################################################
-# Configuration.
+# # Configuration | @version 1.05?
+# You are an intelligent assistant specializing in interviews with business clients
+# for in-depth content creation, etc..()
 ###############################################################################
 
-#if not torch.cuda.is_available():
-    #DESCRIPTION += "\n<p>⚠️Running on CPU, This may not work on CPU.</p>"
+device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
+
+###############################################################################
+# Function to detect leading silence
+###############################################################################
 
-#CACHE_EXAMPLES = torch.device('cuda') and os.getenv("CACHE_EXAMPLES", "0") == "1"
-#CACHE_EXAMPLES = torch.cuda.is_available() and os.getenv("CACHE_EXAMPLES", "0") == "1"
-#USE_TORCH_COMPILE = os.getenv("USE_TORCH_COMPILE", "0") == "1"
-#ENABLE_CPU_OFFLOAD = os.getenv("ENABLE_CPU_OFFLOAD", "0") == "1"
+def milliseconds_until_sound(sound, silence_threshold_in_decibels=-20.0, chunk_size=10):
+    trim_ms = 0
+    assert chunk_size > 0
+    while sound[trim_ms:trim_ms + chunk_size].dBFS < silence_threshold_in_decibels and trim_ms < len(sound):
+        trim_ms += chunk_size
+    return trim_ms
 
-device = torch.device('cuda')
-#device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
+###############################################################################
+# Trim the start of the audio file
+###############################################################################
 
+def trim_start(filepath):
+    path = Path(filepath)
+    directory = path.parent
+    filename = path.name
+    audio = AudioSegment.from_file(filepath, format="wav")
+    start_trim = milliseconds_until_sound(audio)
+    trimmed = audio[start_trim:]
+    new_filename = directory / f"trimmed_{filename}"
+    trimmed.export(new_filename, format="wav")
+    return trimmed, new_filename
 
-def transcribe(file_upload, progress=gr.Progress(track_tqdm=True)): # microphone
+###############################################################################
+# -- segment the audio into smaller parts (1-minute segments for large files)
+###############################################################################
+
+def segment_audio(trimmed_audio, output_dir_trimmed):
+    one_minute = 1 * 60 * 1000  # 1 minute in milliseconds
+    start_time = 0
+    i = 0
+
+    # -- iterate through trimmed audio, segment it
+    segmented_files = []
+    while start_time < len(trimmed_audio):
+        segment = trimmed_audio[start_time:start_time + one_minute]
+
+        # -- filename for each segment
+        file_name = f"trimmed_{i:02d}.wav"
+
+        # --export each segment, save to the Hugging Face hub directly
+        file_path = file_name
+        segment.export(file_path, format="wav")
+
+
+        segmented_files.append(file_path)
+        start_time += one_minute
+        i += 1
+
+    return segmented_files
+
+
+###############################################################################
+# Transcription logic
+###############################################################################
 
-    file = file_upload 
+def transcribe(file_upload, progress=gr.Progress(track_tqdm=True)):
+    file = file_upload
     start_time = time.time()
 
-    #--------------____________________________________________--------------"
-    with torch.no_grad():
-        pipe = pipeline("automatic-speech-recognition", 
-                        model="NbAiLab/nb-whisper-large", 
-                        chunk_length_s=30, 
-                        device=device)
-        
-    pipe.model.config.forced_decoder_ids = pipe.tokenizer.get_decoder_prompt_ids(task="transcribe")
-    
-    text = pipe(file)["text"]
-    #--------------____________________________________________--------------"
+    # -- trim auio, segment it for processing
+    trimmed_audio, trimmed_filename = trim_start(file)
+    segmented_files = segment_audio(trimmed_audio, "trimmed_audio")
+
+
+    pipe = pipeline("automatic-speech-recognition", model="NbAiLab/nb-whisper-large", chunk_length_s=30, device=device)
+
+    transcriptions = [pipe(seg_file)["text"] for seg_file in segmented_files]
+    text = ''.join(transcriptions)
 
     end_time = time.time()
     output_time = end_time - start_time
-    
-    # --word count
+
+    # --Word count
     word_count = len(text.split())
 
-    # --memory metrics
-    memory = psutil.virtual_memory()
-    
-    # --cpu metric
+    # --CPU metric
     cpu_usage = psutil.cpu_percent(interval=1)
 
-    # --gpu metric
-    gpu_utilization, gpu_memory = GPUInfo.gpu_usage()
-    
     # --system info string
     system_info = f"""
     Processing time: {output_time:.2f} seconds.
     Number of words: {word_count}
     CPU Usage: {cpu_usage}%
-    GPU Memory: {gpu_memory}%
-    GPU Utilization: {gpu_utilization}%
     """
 
-    return text.strip(), system_info
+
+    return text, system_info
+
 
 ###############################################################################
-# Interface.
+# Interface
 ###############################################################################
 
 HEADER_INFO = """
-    # SWITCHVOX ✨|🇳🇴 *Transkribering av lydfiler til norsk bokmål.*
+    # SWITCHVOX ✨|🇳🇴 *Transkribering av lydfiler til Norsk skrift.*
 """.strip()
-LOGO = "https://cdn-lfs-us-1.huggingface.co/repos/fe/3b/fe3bd7c8beece8b087fddcc2278295e7f56c794c8dcf728189f4af8bddc585e1/24ad06a03a5bc66f3eba361b94e45ad17e46f98b76632f2d17faf8a0b4f9ab6b?response-content-disposition=inline%3B+filename*%3DUTF-8%27%27banner_trans.png%3B+filename%3D%22banner_trans.png%22%3B&response-content-type=image%2Fpng&Expires=1726757282&Policy=eyJTdGF0ZW1lbnQiOlt7IkNvbmRpdGlvbiI6eyJEYXRlTGVzc1RoYW4iOnsiQVdTOkVwb2NoVGltZSI6MTcyNjc1NzI4Mn19LCJSZXNvdXJjZSI6Imh0dHBzOi8vY2RuLWxmcy11cy0xLmh1Z2dpbmdmYWNlLmNvL3JlcG9zL2ZlLzNiL2ZlM2JkN2M4YmVlY2U4YjA4N2ZkZGNjMjI3ODI5NWU3ZjU2Yzc5NGM4ZGNmNzI4MTg5ZjRhZjhiZGRjNTg1ZTEvMjRhZDA2YTAzYTViYzY2ZjNlYmEzNjFiOTRlNDVhZDE3ZTQ2Zjk4Yjc2NjMyZjJkMTdmYWY4YTBiNGY5YWI2Yj9yZXNwb25zZS1jb250ZW50LWRpc3Bvc2l0aW9uPSomcmVzcG9uc2UtY29udGVudC10eXBlPSoifV19&Signature=HB0ypHpwK3dgXHqU5a3oBoR-OlPTV-Zlo-QzpvVD8DOlYvLCIwheHxh6OFUSFiWt1qEhWaelL71O1Rx5EwHG8L6oKbOVEvvrHzZjIJ9RD2YlOlx96EG5ZlaVdAlT0trDwlre-Q8VVey22UAu-H9hX%7EoyLoksIgbWX02%7E5ncmeujYG0KRMVwwB9DCkOY6FxtISGAw2A7qv1FoOdJ6nMxi8ijXDlmRigY9Cr-iuqYOUCBv4oinK-d-LEljUTbWEua1t8BvvlE02yt1TQGd8xz6E-qzWQN%7Es8%7EjNZRGMybpk5FaIl8%7El%7EMmr2Iy%7Erh62180ffBHG5YUgPnpmDKiKA2P-g__&Key-Pair-Id=K24J24Z295AEI9"
-SIDEBAR_INFO = f"""
-<div align="center">
-    <img src="{LOGO}" style="width: 100%; height: auto;"/>
-</div>
-"""
-
-"""
-def save_to_pdf(text, summary):
-    pdf = FPDF()
-    pdf.add_page()
-    pdf.set_font("Arial", size=12)
-
-    if text:
-        pdf.multi_cell(0, 10, "Transkribert Tekst:\n" + text)
-
-    pdf.ln(10)  # Paragraph metric
-
-    if summary:
-        pdf.multi_cell(0, 10, "Summary:\n" + summary)
-
-    pdf_output_path = "transcription_.pdf"
-    pdf.output(pdf_output_path)
-    return pdf_output_path
-"""
-
 
 css = """
 #transcription_output textarea {
@@ -153,56 +166,36 @@ iface = gr.Blocks(css=css)
 
 with iface:
 
-    gr.HTML(SIDEBAR_INFO)
     gr.Markdown(HEADER_INFO)
 
     with gr.Row():
-        gr.Markdown('''
-        ##### 🔊 Last opp lydfila [max.lengde: 40min]
-        ##### ☕️ Trykk på "Transkriber" knappen og vent på svar
-        ##### ⚡️ Går rimelig bra kjapt med Norwegian NB-Whisper Large..
-        ##### 😅 Planlegger tilleggs-funksjoner senere
-        
-        ''')
-        #microphone = gr.Audio(label="Microphone", sources="microphone", type="filepath")
         upload = gr.Audio(label="Upload audio", sources="upload", type="filepath")
         transcribe_btn = gr.Button("Transkriber")
 
-    with gr.Row():   
+    with gr.Row():
         with gr.Column(scale=3):
-            text_output = gr.Textbox(label="Transkribert Tekst", elem_id="transcription_output")
+            text_output = gr.Textbox(label="Transkribert Tekst", placeholder="t r a n s c r i p t i o", elem_id="transcription_output")
         with gr.Column(scale=1):
             system_info = gr.Textbox(label="Antall sekunder, ord, system data:", elem_id="system_info_box")
-        
-    """
-    with gr.Tabs():
-        with gr.TabItem("Download PDF"):
-            pdf_text_only = gr.Button("Last ned pdf med resultat")
-            pdf_output = gr.File(label="/.pdf")
 
-            pdf_text_only.click(fn=lambda text: save_to_pdf(text, ""), inputs=[text_output], outputs=[pdf_output])
-    """
-    
     with gr.Row():
         gr.Markdown('''
-        <div align="center">
-            <a href="https://opensource.com/resources/what-open-source">
-                <img src="https://badgen.net/badge/Open%20Source%20%3F/Yes%21/blue?icon=github" alt="Open Source? Yes!">
+        <div style="text-align:center;">
+        <a href="https://opensource.com/resources/what-open-source" style="display: inline-block;">
+            <img src="https://badgen.net/badge/Open%20Source%20%3F/Yes%21/blue?icon=github" alt="Open Source? Yes!" style="vertical-align: middle;">
         </a>
-        <span style="display:inline-block; width: 20px;"></span> 
-        <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">
+        <span style="display:inline-block; width: 20px;"></span> <!-- This adds space between the logos -->
+        <a href="https://opensource.org/licenses/Apache-2.0" style="display: inline-block;">
+            <img src="https://img.shields.io/badge/License-Apache_2.0-blue.svg" alt="License: Apache 2.0" style="vertical-align: middle;">
         </a>
         </div>
         ''')
+
+  
     transcribe_btn.click(
-        fn=transcribe, 
-        inputs=[upload], # microphone 
+        fn=transcribe,
+        inputs=[upload],
         outputs=[text_output, system_info]
-    )    
-    
-    #transcribe_btn.click(fn=transcribe, inputs=[microphone, upload], outputs=[text_output, system_info])
-
-
+    )
 
-iface.launch(share=True,debug=True)
+iface.launch(debug=True)