second commit with streamlit

.gitignore

@@ -3,6 +3,7 @@ build
 dist
 speechlib.egg-info
 .env
+.venv
 
 *.swp
 *.swo

README.md

@@ -132,14 +132,6 @@ end: ending time of speech in seconds
 text: transcribed text for speech during start and end  
 speaker: speaker of the text 
 
-#### voices_folder structure:  
-
-![voices_folder_structure](voices_folder_structure1.png)
-
-#### Transcription:  
-
-![transcription](transcript.png)
-
 supported language codes:  
 
 ```

config.ini

@@ -1,11 +1,11 @@
 [FILE]
-mode = cpu
 modelpath = d:\faster-whisper-large-v3-turbo-ct2
 audiofolder = \audios
-outputfolder = D:\speechlib\output
+outputfolder = D:\Pythoncode\speechlib\output
-outputfolder2 = D:\speechlib\output
+outputfolder2 = D:\Pythoncode\speechlib\output
-outputfolder3 = D:\speechlib\output
+outputfolder3 = D:\Pythoncode\speechlib\output
-voicefolder = D:\speechlib\voices
+voicefolder = D:\Pythoncode\speechlib\voices
-beamsize = 6
+beamsize = 5
-batchsize = 8
+batchsize = 4
-accesstoken = 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
+accesstoken = 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
+quantization = False

main.py

@@ -1,16 +1,17 @@
 from speechlib import Transcriptor
-from speechlib import PreProcessor
 from configparser import ConfigParser
 import os
 
+def str2bool(v):
+  return v.lower() in ("yes", "true", "t", "1")
+
 config = ConfigParser()
 config.read('config.ini')
-mode = config.get('FILE', 'mode')
 audiofolder = config.get('FILE', 'audiofolder')
 access_token = config.get('FILE', 'accesstoken')
 voicefolder = config.get('FILE', 'voicefolder')
 language = "id"
-quantization = False
+quantization = str2bool(config.get('FILE', 'quantization'))
 modelSize = "medium"
 
 ### load the audio file in audio folder ###
@@ -27,7 +28,7 @@ if not os.path.exists(output_dir):
     
 ### Loop for each audio file in the audio folder ###
 for filename in os.listdir(audio_dir):
-    if filename.endswith(".mp3") or filename.endswith(".wav"):
+    if filename.lower().endswith(".mp3") or filename.lower().endswith(".wav"):
         audiofile = os.path.join(audio_dir, filename)
         print(f"Audio file: {audiofile}")
         audiofilewithoutextension = audiofile.split(".mp3")[0].split(".wav")[0]

main2.py

@@ -0,0 +1,57 @@
+from speechlib import Transcriptor
+from configparser import ConfigParser
+import os
+import streamlit as st
+import tempfile
+
+os.environ["KMP_DUPLICATE_LIB_OK"]="TRUE"
+
+def str2bool(v):
+  return v.lower() in ("yes", "true", "t", "1")
+
+config = ConfigParser()
+config.read('config.ini')
+audiofolder = config.get('FILE', 'audiofolder')
+access_token = config.get('FILE', 'accesstoken')
+voicefolder = config.get('FILE', 'voicefolder')
+outputfolder = config.get('FILE', 'outputfolder')
+language = "id"
+quantization = str2bool(config.get('FILE', 'quantization'))
+modelSize = "medium"
+
+st.title("Crew Transcription")
+
+audio = st.file_uploader("Upload audio file (.mp3/.wav/.m4a)", type=["mp3", "wav", "m4a"])
+crew_id = st.text_input("Crew ID")
+crew_name = st.text_input("Crew Name")
+outlet = st.selectbox('Outlet',('Icon Mall','Kediri','Ponti','Plaza Surabaya Delta','AEON','Rungkut','Darmo','Royal Plaza',
+         'Kusuma Bangsa','Sepanjang','Barata','Pakuwon City Mall','Tropodo','Wiyung','Citraland',
+         'Lotte Avenue','Pakuwon Trade Center','Rest Area Sidoarjo','Gayungsari','Wahidin','Imam Bonjol',
+         'BG Junction','Puri Surya Jaya','Merr','Karawaci','Pepelegi','Plaza Festival'
+         ))
+date = st.date_input("Date")
+if st.button("Transcribe"):
+    if audio is not None and crew_id:
+        with st.spinner("Transcribing..."):
+            # Save the uploaded audio file to a temporary location
+            file_details = {"filename": audio.name, "filetype": audio.type, "filesize": audio.size}
+            print(file_details)
+            
+            temp_dir = tempfile.mkdtemp()
+            path = os.path.join(temp_dir, audio.name)
+            
+            with open(path, "wb") as f:
+                f.write(audio.getvalue())
+            print(f"Temporary file saved at: {path}")
+            
+            #audiofilewithoutextension = path.split(".mp3")[0].split(".wav")[0].split(".m4a")[0]
+            #filepath = os.path.join(outputfolder, os.path.basename(audiofilewithoutextension).split('/')[-1]+".txt")
+            filepath = os.path.join(outputfolder, f"{outlet}-{crew_id}-{crew_name}-{date}-Transkrip.txt")
+            print(f"Output file: {filepath}")
+            filename = open(filepath, "w")
+
+            ### transcribe ###
+            transcriptor = Transcriptor(path, filepath, language, modelSize, access_token, voicefolder, quantization)
+            res = transcriptor.faster_whisper()
+            print(f"Content has been written to {filepath}")
+            st.success(f"Transcribe successful!")

pretrained_models/spkrec-ecapa-voxceleb/classifier.ckpt

@@ -0,0 +1 @@
+C:/Users/admin.suherdy/.cache/huggingface/hub/models--speechbrain--spkrec-ecapa-voxceleb/snapshots/0f99f2d0ebe89ac095bcc5903c4dd8f72b367286/classifier.ckpt

pretrained_models/spkrec-ecapa-voxceleb/embedding_model.ckpt

@@ -0,0 +1 @@
+C:/Users/admin.suherdy/.cache/huggingface/hub/models--speechbrain--spkrec-ecapa-voxceleb/snapshots/0f99f2d0ebe89ac095bcc5903c4dd8f72b367286/embedding_model.ckpt

pretrained_models/spkrec-ecapa-voxceleb/hyperparams.yaml

@@ -1,58 +0,0 @@
-# ############################################################################
-# Model: ECAPA big for Speaker verification
-# ############################################################################
-
-# Feature parameters
-n_mels: 80
-
-# Pretrain folder (HuggingFace)
-pretrained_path: speechbrain/spkrec-ecapa-voxceleb
-
-# Output parameters
-out_n_neurons: 7205
-
-# Model params
-compute_features: !new:speechbrain.lobes.features.Fbank
-    n_mels: !ref <n_mels>
-
-mean_var_norm: !new:speechbrain.processing.features.InputNormalization
-    norm_type: sentence
-    std_norm: False
-
-embedding_model: !new:speechbrain.lobes.models.ECAPA_TDNN.ECAPA_TDNN
-    input_size: !ref <n_mels>
-    channels: [1024, 1024, 1024, 1024, 3072]
-    kernel_sizes: [5, 3, 3, 3, 1]
-    dilations: [1, 2, 3, 4, 1]
-    attention_channels: 128
-    lin_neurons: 192
-
-classifier: !new:speechbrain.lobes.models.ECAPA_TDNN.Classifier
-    input_size: 192
-    out_neurons: !ref <out_n_neurons>
-
-mean_var_norm_emb: !new:speechbrain.processing.features.InputNormalization
-    norm_type: global
-    std_norm: False
-
-modules:
-    compute_features: !ref <compute_features>
-    mean_var_norm: !ref <mean_var_norm>
-    embedding_model: !ref <embedding_model>
-    mean_var_norm_emb: !ref <mean_var_norm_emb>
-    classifier: !ref <classifier>
-        
-label_encoder: !new:speechbrain.dataio.encoder.CategoricalEncoder
-
-        
-pretrainer: !new:speechbrain.utils.parameter_transfer.Pretrainer
-    loadables:
-        embedding_model: !ref <embedding_model>
-        mean_var_norm_emb: !ref <mean_var_norm_emb>
-        classifier: !ref <classifier>
-        label_encoder: !ref <label_encoder>
-    paths:
-        embedding_model: !ref <pretrained_path>/embedding_model.ckpt
-        mean_var_norm_emb: !ref <pretrained_path>/mean_var_norm_emb.ckpt
-        classifier: !ref <pretrained_path>/classifier.ckpt
-        label_encoder: !ref <pretrained_path>/label_encoder.txt

pretrained_models/spkrec-ecapa-voxceleb/hyperparams.yaml

@@ -0,0 +1 @@
+C:/Users/admin.suherdy/.cache/huggingface/hub/models--speechbrain--spkrec-ecapa-voxceleb/snapshots/0f99f2d0ebe89ac095bcc5903c4dd8f72b367286/hyperparams.yaml

pretrained_models/spkrec-ecapa-voxceleb/label_encoder.ckpt

@@ -0,0 +1 @@
+C:/Users/admin.suherdy/.cache/huggingface/hub/models--speechbrain--spkrec-ecapa-voxceleb/snapshots/0f99f2d0ebe89ac095bcc5903c4dd8f72b367286/label_encoder.txt

pretrained_models/spkrec-ecapa-voxceleb/mean_var_norm_emb.ckpt

@@ -0,0 +1 @@
+C:/Users/admin.suherdy/.cache/huggingface/hub/models--speechbrain--spkrec-ecapa-voxceleb/snapshots/0f99f2d0ebe89ac095bcc5903c4dd8f72b367286/mean_var_norm_emb.ckpt

requirements.txt

@@ -5,3 +5,7 @@ speechbrain
 accelerate
 faster-whisper
 openai-whisper
+streamlit
+torch
+torchaudio
+assemblyai

run1.bat

@@ -0,0 +1 @@
+streamlit run main2.py --server.headless true --server.maxUploadSize 1000

run2.bat

@@ -0,0 +1 @@
+streamlit run main3.py --server.headless true --server.maxUploadSize 1000

run3.bat

@@ -0,0 +1 @@
+streamlit run main4.py --server.headless true --server.maxUploadSize 1000

speechlib/core_analysis.py

@@ -11,12 +11,16 @@ from .re_encode import (re_encode)
 from .convert_to_mono import (convert_to_mono)
 from .convert_to_wav import (convert_to_wav)
 
+import subprocess
+
 # by default use google speech-to-text API
 # if False, then use whisper finetuned version for sinhala
 def core_analysis(file_name, voices_folder, log_folder, language, modelSize, ACCESS_TOKEN, model_type, quantization=False, custom_model_path=None, hf_model_id=None, aai_api_key=None):
 
     # <-------------------PreProcessing file-------------------------->
-
+    # convert compressed wav
+    # if file_name.lower().endswith(".wav"):
+    #     subprocess.call(['ffmpeg','i',file_name,file_name.split(".")[0]+'.wav'])
     # check if file is in wav format, if not convert to wav
     file_name = convert_to_wav(file_name)
 
@@ -136,4 +140,6 @@ def core_analysis(file_name, voices_folder, log_folder, language, modelSize, ACC
     # writing log file
     write_log_file(common_segments, log_folder, file_name, language)  
 
+    ## TODO cleaning segments and temp folder
+
     return common_segments

speechlib/speaker_recognition.py

@@ -38,7 +38,7 @@ def speaker_recognition(file_name, voices_folder, segments, wildcards):
         end = segment[1] * 1000     # end time in miliseconds
         clip = audio[start:end]
         i = i + 1
-        file = folder_name + "/" + file_name.split("/")[-1].split(".")[0] + "_segment"+ str(i) + ".wav"
+        file = folder_name + "/" + file_name.split("\\")[-1].split(".")[0] + "_segment"+ str(i) + ".wav"
         clip.export(file, format="wav")
 
         max_score = 0
@@ -69,7 +69,10 @@ def speaker_recognition(file_name, voices_folder, segments, wildcards):
         Id_count[person] += 1
 
         # Delete the WAV file after processing
-        os.remove(file)
+        # try:
+        #     os.remove(file)
+        # except OSError as e:
+        #     print (f'Access-error on file {str(e)}')
 
         current_pred = max(Id_count, key=Id_count.get)
 

speechlib/wav_segmenter.py

@@ -23,7 +23,7 @@ def wav_file_segmentation(file_name, segments, language, modelSize, model_type,
         end = segment[1] * 1000     # end time in miliseconds
         clip = audio[start:end]
         i = i + 1
-        file = folder_name + "/" + f"segment-{file_name}"+ str(i) + ".wav"
+        file = folder_name + "/" + f"segment-{file_name.split("\\")[-1].split(".")[0]}"+ str(i) + ".wav"
         clip.export(file, format="wav")
 
         try:
@@ -34,10 +34,10 @@ def wav_file_segmentation(file_name, segments, language, modelSize, model_type,
         except Exception as err:
             print("ERROR while transcribing: ", err)
         # Delete the WAV file after processing
-        try:
+        # try:
-            os.remove(file)
+        #     os.remove(file)
-        except OSError as e:
+        # except OSError as e:
-            print (f'Access-error on file {str(e)}')
+        #     print (f'Access-error on file {str(e)}')
         
 
     return texts