some corrections

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 interent 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. 
 
 ### Instructions 
 #### Requirements
@@ -22,7 +24,7 @@ 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).
 
 ### Example
-See the [example](example.ipynb) implementation on 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. 
 
 [^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.
 

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
+}

transcribe.py

@@ -17,11 +17,10 @@ def transcribe(path, file_type, model=None, language=None, verbose=True):
     print('Expecting {} language, you can change this by specifying language="English". None will try to auto-detect'.format(language))
     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))
     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]