Update transcribe.py

better time keeping

CITATION.cff

@@ -1,11 +1,11 @@
 cff-version: 1.2.0
-message: "If you use find this implementation useful in your research, please cite it as below."
+message: "If you use find this implementation useful in your research, and want to cite it, please do so as below."
 authors:
 - family-names: "Söderström"
   given-names: "Kristofer Rolf"
   orcid: "https://orcid.org/0000-0002-5322-3350"
 title: "transcribe"
-version: 1.0
+version: 1.1.1
-doi: None
+doi: 10.5281/zenodo.7760511
 date-released: 2023-03-22
 url: "https://github.com/soderstromkr/transcribe"

GUI.py

@@ -0,0 +1,100 @@
+import tkinter as tk
+from tkinter import ttk
+from tkinter import filedialog
+from tkinter import messagebox
+from transcribe import transcribe
+from ttkthemes import ThemedTk
+import whisper 
+import numpy as np
+import glob, os
+
+
+class App:
+    def __init__(self, master):
+        self.master = master
+        master.title("Local Transcribe")
+
+        #style options
+        style = ttk.Style()
+        style.configure('TLabel', font=('Arial', 10), padding=10)
+        style.configure('TEntry', font=('Arial', 10), padding=10)
+        style.configure('TButton', font=('Arial', 10), padding=10)
+        style.configure('TCheckbutton', font=('Arial', 10), padding=10)
+
+        # Folder Path
+        path_frame = ttk.Frame(master, padding=10)
+        path_frame.pack(fill=tk.BOTH)
+        path_label = ttk.Label(path_frame, text="Folder Path:")
+        path_label.pack(side=tk.LEFT, padx=5)
+        self.path_entry = ttk.Entry(path_frame, width=50)
+        self.path_entry.insert(10, 'sample_audio/')
+        self.path_entry.pack(side=tk.LEFT, fill=tk.X, expand=True)
+        browse_button = ttk.Button(path_frame, text="Browse", command=self.browse)
+        browse_button.pack(side=tk.LEFT, padx=5)
+
+        # File Type
+        file_type_frame = ttk.Frame(master, padding=10)
+        file_type_frame.pack(fill=tk.BOTH)
+        file_type_label = ttk.Label(file_type_frame, text="File Type:")
+        file_type_label.pack(side=tk.LEFT, padx=5)
+        self.file_type_entry = ttk.Entry(file_type_frame, width=50)
+        self.file_type_entry.insert(10, 'ogg')
+        self.file_type_entry.pack(side=tk.LEFT, fill=tk.X, expand=True)
+
+        # Model
+        model_frame = ttk.Frame(master, padding=10)
+        model_frame.pack(fill=tk.BOTH)
+        model_label = ttk.Label(model_frame, text="Model:")
+        model_label.pack(side=tk.LEFT, padx=5)
+        self.model_entry = ttk.Entry(model_frame, width=50)
+        self.model_entry.insert(10, 'small')
+        self.model_entry.pack(side=tk.LEFT, fill=tk.X, expand=True)
+
+        # Language (currently disabled)
+        #language_frame = ttk.Frame(master, padding=10)
+        #language_frame.pack(fill=tk.BOTH)
+        #language_label = ttk.Label(language_frame, text="Language:")
+        #language_label.pack(side=tk.LEFT, padx=5)
+        #self.language_entry = ttk.Entry(language_frame, width=50)
+        #self.language_entry.insert(10, np.nan)
+        #self.language_entry.pack(side=tk.LEFT, fill=tk.X, expand=True)
+
+        # Verbose
+        verbose_frame = ttk.Frame(master, padding=10)
+        verbose_frame.pack(fill=tk.BOTH)
+        self.verbose_var = tk.BooleanVar()
+        verbose_checkbutton = ttk.Checkbutton(verbose_frame, text="Verbose", variable=self.verbose_var)
+        verbose_checkbutton.pack(side=tk.LEFT, padx=5)
+
+        # Buttons
+        button_frame = ttk.Frame(master, padding=10)
+        button_frame.pack(fill=tk.BOTH)
+        transcribe_button = ttk.Button(button_frame, text="Transcribe Audio", command=self.transcribe)
+        transcribe_button.pack(side=tk.LEFT, padx=5, pady=10, fill=tk.X, expand=True)
+        quit_button = ttk.Button(button_frame, text="Quit", command=master.quit)
+        quit_button.pack(side=tk.RIGHT, padx=5, pady=10, fill=tk.X, expand=True)
+
+    def browse(self):
+        folder_path = filedialog.askdirectory()
+        self.path_entry.delete(0, tk.END)
+        self.path_entry.insert(0, folder_path)
+
+    def transcribe(self):
+        path = self.path_entry.get()
+        file_type = self.file_type_entry.get()
+        model = self.model_entry.get()
+        #language = self.language_entry.get()
+        language = None # set to auto-detect
+        verbose = self.verbose_var.get()
+
+        # Call the transcribe function with the appropriate arguments
+        result = transcribe(path, file_type, model=model, language=language, verbose=verbose)
+
+        # Show the result in a message box
+        tk.messagebox.showinfo("Finished!", result)
+
+if __name__ == "__main__":
+#    root = tk.Tk()
+    root = ThemedTk(theme="clearlooks")
+    app = App(root) 
+    root.mainloop()

Mac_instructions.txt

@@ -0,0 +1,5 @@
+### How to run on Mac
+Unfortunately, I have not found a permament solution for this, not being a Mac user has limited the ways I can test this. For now, these are the recommended steps for a beginner user:
+1. Open a terminal and navigate to the root folder (transcribe-main if you downloaded the folder). You can also right-click (or equivalent) on the root folder to open a Terminal within the folder.
+2. Run the following command: 
+python GUI.py

README.md

@@ -1,5 +1,7 @@
 ## transcribe
 Simple script that uses OpenAI's Whisper to transcribe audio files from your local folders. 
+## Note
+This implementation and guide is mostly made for researchers not familiar with programming that want a way to transcribe their files locally, without internet connection, usually required within ethical data practices and frameworks. Two examples are shown, a normal workflow with internet connection. And one in which the model is loaded first, via openai-whisper, and then the transcription can be done without being connected to the internet. There is now also a GUI implementation, read below for more information.  
 
 ### Instructions 
 #### Requirements
@@ -14,14 +16,28 @@ Users might not need to specifically install Transfomers. However, a conda insta
 ```
 pip install -U openai-whisper
 ```
+4. There is an option to run a batch file, which launches a GUI built on TKinter and TTKthemes. If using these options, make sure they are installed in your python build. You can install them via pip.
+```
+pip install tk
+```
+and
+```
+pip install ttkthemes 
+```
 #### Using the script
 This is a simple script with no installation. You can either clone the repository with
 ```
 git clone https://github.com/soderstromkr/transcribe.git
 ```
-and use the example.ipynb template to use the script **OR (for beginners)** download the ```transcribe.py``` file into your work folder. Then you can either import it to another script or notebook for use. I recommend jupyter notebook for new users, see the example below. (Remember to have transcribe.py and example.ipynb in the same working folder).
+and use the example.ipynb template to use the script.
+**OR** download the ```transcribe.py``` file into your work folder. Then you can either import it to another script or notebook for use. I recommend jupyter notebook for new users, see the example below. (Remember to have transcribe.py and example.ipynb in the same working folder). 
+#### Example with jupyter notebook 
+See [example](example.ipynb) for an implementation on jupyter notebook, also added an example for a simple [workaround](example_no_internet.ipynb) to transcribe while offline. 
+#### Using the GUI 
+You can also run the GUI version from your terminal running ```python GUI.py``` or with the batch file called run_Windows.bat (for Windows user, Mac users should read the text file for instructions), just make sure to add your conda path to it. If you want to download a model first, and then go offline for transcription, I recommend running the model with the default sample folder, which will download the model locally. The GUI should look like this:  
+![python GUI.py](gui_jpeg.jpg?raw=true)
 
-### Example
-See the [example](example.ipynb) implementation on jupyter notebook. 
 
 [^1]: Advanced users can use ```pip install ffmpeg-python``` but be ready to deal with some [PATH issues](https://stackoverflow.com/questions/65836756/python-ffmpeg-wont-accept-path-why), which I encountered in Windows 11.
+
+[![DOI](https://zenodo.org/badge/617404576.svg)](https://zenodo.org/badge/latestdoi/617404576)

example.ipynb

@@ -40,7 +40,7 @@
    "outputs": [],
    "source": [
     "path='sample_audio/'#folder path\n",
-    "file_type='ogg' #check your file for file type, will only transcribe files with the file type, 'ogg', 'WAV'\n",
+    "file_type='ogg' #check your file for file type, will only transcribe those files\n",
     "model='medium' #'small', 'medium', 'large' (tradeoff between speed and accuracy)\n",
     "language= None #tries to auto-detect, other options include 'English', 'Spanish', etc...\n",
     "verbose = True # prints output while transcribing, False to deactivate"

example_no_internet.ipynb

@@ -0,0 +1,231 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "id": "eba9e610",
+   "metadata": {},
+   "source": [
+    "A simple way to avoid being connected while transcribing is to first load the model version you want to use. See [here](https://github.com/openai/whisper/blob/main/README.md#available-models-and-languages) for more info."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 6,
+   "id": "85cd2d12",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "Whisper(\n",
+       "  (encoder): AudioEncoder(\n",
+       "    (conv1): Conv1d(80, 1024, kernel_size=(3,), stride=(1,), padding=(1,))\n",
+       "    (conv2): Conv1d(1024, 1024, kernel_size=(3,), stride=(2,), padding=(1,))\n",
+       "    (blocks): ModuleList(\n",
+       "      (0-23): 24 x ResidualAttentionBlock(\n",
+       "        (attn): MultiHeadAttention(\n",
+       "          (query): Linear(in_features=1024, out_features=1024, bias=True)\n",
+       "          (key): Linear(in_features=1024, out_features=1024, bias=False)\n",
+       "          (value): Linear(in_features=1024, out_features=1024, bias=True)\n",
+       "          (out): Linear(in_features=1024, out_features=1024, bias=True)\n",
+       "        )\n",
+       "        (attn_ln): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)\n",
+       "        (mlp): Sequential(\n",
+       "          (0): Linear(in_features=1024, out_features=4096, bias=True)\n",
+       "          (1): GELU(approximate='none')\n",
+       "          (2): Linear(in_features=4096, out_features=1024, bias=True)\n",
+       "        )\n",
+       "        (mlp_ln): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)\n",
+       "      )\n",
+       "    )\n",
+       "    (ln_post): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)\n",
+       "  )\n",
+       "  (decoder): TextDecoder(\n",
+       "    (token_embedding): Embedding(51865, 1024)\n",
+       "    (blocks): ModuleList(\n",
+       "      (0-23): 24 x ResidualAttentionBlock(\n",
+       "        (attn): MultiHeadAttention(\n",
+       "          (query): Linear(in_features=1024, out_features=1024, bias=True)\n",
+       "          (key): Linear(in_features=1024, out_features=1024, bias=False)\n",
+       "          (value): Linear(in_features=1024, out_features=1024, bias=True)\n",
+       "          (out): Linear(in_features=1024, out_features=1024, bias=True)\n",
+       "        )\n",
+       "        (attn_ln): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)\n",
+       "        (cross_attn): MultiHeadAttention(\n",
+       "          (query): Linear(in_features=1024, out_features=1024, bias=True)\n",
+       "          (key): Linear(in_features=1024, out_features=1024, bias=False)\n",
+       "          (value): Linear(in_features=1024, out_features=1024, bias=True)\n",
+       "          (out): Linear(in_features=1024, out_features=1024, bias=True)\n",
+       "        )\n",
+       "        (cross_attn_ln): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)\n",
+       "        (mlp): Sequential(\n",
+       "          (0): Linear(in_features=1024, out_features=4096, bias=True)\n",
+       "          (1): GELU(approximate='none')\n",
+       "          (2): Linear(in_features=4096, out_features=1024, bias=True)\n",
+       "        )\n",
+       "        (mlp_ln): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)\n",
+       "      )\n",
+       "    )\n",
+       "    (ln): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)\n",
+       "  )\n",
+       ")"
+      ]
+     },
+     "execution_count": 6,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "import whisper\n",
+    "#change to model size, bigger is more accurate but slower\n",
+    "whisper.load_model(\"medium\") #base, small, medium, large"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 7,
+   "id": "0d2acd54",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "#after it loads, you can disconnect from the internet and run the rest"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 8,
+   "id": "a2cd4050",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from transcribe import transcribe"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 9,
+   "id": "24e1d24e",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Help on function transcribe in module transcribe:\n",
+      "\n",
+      "transcribe(path, file_type, model=None, language=None, verbose=True)\n",
+      "    Implementation of OpenAI's whisper model. Downloads model, transcribes audio files in a folder and returns the text files with transcriptions\n",
+      "\n"
+     ]
+    }
+   ],
+   "source": [
+    "help(transcribe)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 11,
+   "id": "e52477fb",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "path='sample_audio/'#folder path\n",
+    "file_type='ogg' #check your file for file type, will only transcribe those files\n",
+    "model='medium' #'small', 'medium', 'large' (tradeoff between speed and accuracy)\n",
+    "language= None #tries to auto-detect, other options include 'English', 'Spanish', etc...\n",
+    "verbose = True # prints output while transcribing, False to deactivate"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 12,
+   "id": "d66866af",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Using medium model, you can change this by specifying model=\"medium\" for example\n",
+      "Only looking for file type ogg, you can change this by specifying file_type=\"mp3\"\n",
+      "Expecting None language, you can change this by specifying language=\"English\". None will try to auto-detect\n",
+      "Verbosity is True. If TRUE it will print out the text as it is transcribed, you can turn this off by setting verbose=False\n",
+      "\n",
+      "There are 2 ogg files in path: sample_audio/\n",
+      "\n",
+      "\n",
+      "Loading model...\n",
+      "Transcribing file number number 1: Armstrong_Small_Step\n",
+      "Model and file loaded...\n",
+      "Starting transcription...\n",
+      "\n",
+      "Detecting language using up to the first 30 seconds. Use `--language` to specify the language\n",
+      "Detected language: English\n",
+      "[00:00.000 --> 00:24.000]  That's one small step for man, one giant leap for mankind.\n",
+      "\n",
+      "Finished file number 1.\n",
+      "\n",
+      "\n",
+      "\n",
+      "Transcribing file number number 2: Axel_Pettersson_röstinspelning\n",
+      "Model and file loaded...\n",
+      "Starting transcription...\n",
+      "\n",
+      "Detecting language using up to the first 30 seconds. Use `--language` to specify the language\n",
+      "Detected language: Swedish\n",
+      "[00:00.000 --> 00:16.000]  Hej, jag heter Axel Pettersson, jag föddes i Örebro 1976. Jag har varit Wikipedia sen 2008 och jag har översatt röstintroduktionsprojektet till svenska.\n",
+      "\n",
+      "Finished file number 2.\n",
+      "\n",
+      "\n",
+      "\n"
+     ]
+    },
+    {
+     "data": {
+      "text/plain": [
+       "'Finished transcription, files can be found in sample_audio/transcriptions'"
+      ]
+     },
+     "execution_count": 12,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "transcribe(path, file_type, model, language, verbose)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "0bc67265",
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3 (ipykernel)",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.10.4"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

run_Windows.bat

@@ -0,0 +1,5 @@
+@echo off
+echo Starting...
+call conda activate base
+REM OPTION 2 : (KEEP TEXT WITHIN QUOTES AND CHANGE USERNAME) "C:/Users/user/Anaconda3/condabin/activate.bat"
+call python GUI.py

sample_audio/transcriptions/Armstrong_Small_Step.txt

@@ -1,3 +1,5 @@
 Armstrong_Small_Step
 In seconds:
-[0.00 --> 24.00]: That's one small step for man, one giant leap for mankind.
+[0.00 --> 7.00]: I'm going to step off the limb now.
+[7.00 --> 18.00]: That's one small step for man.
+[18.00 --> 24.00]: One giant leap for mankind.

sample_audio/transcriptions/Axel_Pettersson_röstinspelning.txt

@@ -1,3 +1,4 @@
 Axel_Pettersson_röstinspelning
 In seconds:
-[0.00 --> 16.00]: Hej, jag heter Axel Pettersson, jag föddes i Örebro 1976. Jag har varit Wikipedia sen 2008 och jag har översatt röstintroduktionsprojektet till svenska.
+[0.00 --> 6.14]: Hej, jag heter Axel Pettersson. Jag följer bror 1976.
+[6.40 --> 15.10]: Jag har varit vikerpedjan sen 2008 och jag har översatt röstintroduktionsprojektet till svenska.

transcribe.py

@@ -1,27 +1,32 @@
 import whisper 
 import glob, os
+#import torch #uncomment if using torch with cuda, below too
+import datetime
 
-def transcribe(path, file_type, model=None, language=None, verbose=True):
+def transcribe(path, file_type, model=None, language=None, verbose=False):
     '''Implementation of OpenAI's whisper model. Downloads model, transcribes audio files in a folder and returns the text files with transcriptions'''
 
     try:
-        os.mkdir('{}transcriptions'.format(path))
+        os.mkdir('{}/transcriptions'.format(path))
     except FileExistsError:
         pass
     
     glob_file = glob.glob(path+'/*{}'.format(file_type))
-    path = path
 
-    print('Using {} model, you can change this by specifying model="medium" for example'.format(model))
+    #if torch.cuda.is_available():
-    print('Only looking for file type {}, you can change this by specifying file_type="mp3"'.format(file_type))    
+    #    generator = torch.Generator('cuda').manual_seed(42)
-    print('Expecting {} language, you can change this by specifying language="English". None will try to auto-detect'.format(language))
+    #else:
-    print('Verbosity is {}. If TRUE it will print out the text as it is transcribed, you can turn this off by setting verbose=False'.format(verbose))
+    #    generator = torch.Generator().manual_seed(42)
+        
+    print('Using {} model'.format(model))
+    print('File type is {}'.format(file_type))    
+    print('Language is being detected automatically for each file')
+    print('Verbosity is set to {}'.format(verbose))
     print('\nThere are {} {} files in path: {}\n\n'.format(len(glob_file), file_type, path))
+    
     print('Loading model...')
     model = whisper.load_model(model)
         
-    
-    
     for idx,file in enumerate(glob_file):
         title = os.path.basename(file).split('.')[0]
         
@@ -30,22 +35,22 @@ def transcribe(path, file_type, model=None, language=None, verbose=True):
         result = model.transcribe(
             file, 
             language=language, 
-            verbose=True
+            verbose=verbose
         )
         start=[]
         end=[]
         text=[]
         for i in range(len(result['segments'])):
-            start.append(result['segments'][i]['start'])
+            start.append(str(datetime.timedelta(seconds=(result['segments'][i]['start']))))
-            end.append(result['segments'][i]['end'])
+            end.append(str(datetime.timedelta(seconds=(result['segments'][i]['end']))))
             text.append(result['segments'][i]['text'])
         
-        with open("{}transcriptions/{}.txt".format(path,title), 'w', encoding='utf-8') as file:
+        with open("{}/transcriptions/{}.txt".format(path,title), 'w', encoding='utf-8') as file:
             file.write(title)
             file.write('\nIn seconds:')
             for i in range(len(result['segments'])):
-                file.writelines('\n[{:.2f} --> {:.2f}]:{}'.format(start[i], end[i], text[i]))
+                file.writelines('\n[{} --> {}]:{}'.format(start[i], end[i], text[i]))
                 
         print('\nFinished file number {}.\n\n\n'.format(idx+1))
 
-    return 'Finished transcription, files can be found in {}transcriptions'.format(path)    
+    return 'Finished transcription, files can be found in {}/transcriptions'.format(path)