kobim/whisper-local-transcribe
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Revamp: embedded console, faster-whisper, simplified install

Browse Source
This commit is contained in:
Kristofer Söderström
2026-03-02 17:02:16 +01:00
parent 7d3fe1ba26
commit f8cf42733d
12 changed files with 449 additions and 139 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
.gitignore vendored
View File

@@ -7,6 +7,7 @@ __pycache__/
venv/
env/
ENV/
.venv/
# IDE
.vscode/

36
Mac_instructions.md
View File

@@ -1,9 +1,31 @@
### 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.
#### Recommended steps
1. Open a terminal and navigate to the root folder (the downloaded the folder).
1. You can also right-click (or equivalent) on the root folder to open a Terminal within the folder.
2. Run the following command:
### How to run on Mac / Linux
#### Quick start
1. Open Terminal and navigate to the project folder (or right-click the folder and select "Open in Terminal").
2. Make the script executable (only needed once):
```
python app.py
chmod +x run_Mac.sh
```
3. Run it:
```
./run_Mac.sh
```
This will automatically:
- Create a virtual environment (`.venv`)
- Install all dependencies (no admin rights needed)
- Launch the app
#### Manual steps (alternative)
If you prefer to do it manually:
```
python3 -m venv .venv
.venv/bin/python install.py
.venv/bin/python app.py
```
#### Notes
- **Python 3.10+** is required. macOS users can install it from [python.org](https://www.python.org/downloads/) or via `brew install python`.
- **No FFmpeg install needed** — audio decoding is bundled.
- **GPU acceleration** is not available on macOS (Apple Silicon MPS is not supported by CTranslate2). CPU with int8 quantization is still fast.
- On Apple Silicon (M1/M2/M3/M4), the `small` or `base` models run well. `medium` works but is slower.

92
README.md
View File

@@ -1,18 +1,24 @@
## Local Transcribe with Whisper
Local Transcribe with Whisper is a user-friendly desktop application that allows you to transcribe audio and video files using the Whisper ASR system. This application provides a graphical user interface (GUI) built with Python and the Tkinter library, making it easy to use even for those not familiar with programming.
## New in version 1.2!
1. Simpler usage:
1. File type: You no longer need to specify file type. The program will only transcribe elligible files.
2. Language: Added option to specify language, which might help in some cases. Clear the default text to run automatic language recognition.
3. Model selection: Now a dropdown option that includes most models for typical use.
2. New and improved GUI.
![python GUI.py](images/gui-windows.png)
> **⚠ Note for Mac users (Apple Silicon):** This version uses `faster-whisper` (CTranslate2), which does **not** support Apple M-chip GPU acceleration. Transcription will run on CPU, which is slower than OpenAI's Whisper with Metal/CoreML support. The trade-off is a much simpler installation — no conda, no PyTorch, no admin rights. If you'd prefer M-chip GPU acceleration and don't mind a more involved setup, switch to the **classic** release:
> ```
> git checkout classic
> ```
Local Transcribe with Whisper is a user-friendly desktop application that allows you to transcribe audio and video files using the Whisper ASR system, powered by [faster-whisper](https://github.com/SYSTRAN/faster-whisper) (CTranslate2). This application provides a graphical user interface (GUI) built with Python and the Tkinter library, making it easy to use even for those not familiar with programming.
## New in version 2.0!
1. **Switched to faster-whisper** — up to 4× faster transcription with lower memory usage.
2. **No separate FFmpeg installation needed** — audio decoding is handled by the bundled PyAV library.
3. **No admin rights required** — a plain `pip install` covers everything.
4. **No PyTorch dependency** — dramatically smaller install footprint.
5. **`tiny` model added** — smallest and fastest option for quick drafts.
## Features
* Select the folder containing the audio or video files you want to transcribe. Tested with m4a video.
* Choose the language of the files you are transcribing. You can either select a specific language or let the application automatically detect the language.
* Select the Whisper model to use for the transcription. Available models include "base.en", "base", "small.en", "small", "medium.en", "medium", and "large". Models with .en ending are better if you're transcribing English, especially the base and small models.
* Select the Whisper model to use for the transcription. Available models include "tiny", "tiny.en", "base", "base.en", "small", "small.en", "medium", "medium.en", "large-v2", and "large-v3". Models with .en ending are better if you're transcribing English, especially the base and small models.
* **Swedish-optimised models** — [KB-Whisper](https://huggingface.co/collections/KBLab/kb-whisper) from the National Library of Sweden (KBLab) is available in all sizes (tiny → large). These models reduce Word Error Rate by up to 47 % compared to OpenAI Whisper on Swedish speech. The language is set to Swedish automatically when a KB model is selected.
* Enable the verbose mode to receive detailed information during the transcription process.
* Monitor the progress of the transcription with the progress bar and terminal.
* Confirmation dialog before starting the transcription to ensure you have selected the correct folder.
@@ -27,66 +33,58 @@ Or by cloning the repository with:
git clone https://github.com/soderstromkr/transcribe.git
```
### Python Version **(any platform including Mac users)**
1. This script was made and tested in an Anaconda environment with Python 3.10. I recommend miniconda for a smaller installation, and if you're not familiar with Python.
See [here](https://docs.anaconda.com/free/miniconda/miniconda-install/) for instructions. You will **need administrator rights**.
2. Whisper also requires some additional libraries. The [setup](https://github.com/openai/whisper#setup) page states: "The codebase also depends on a few Python packages, most notably HuggingFace Transformers for their fast tokenizer implementation and ffmpeg-python for reading audio files."
Users might not need to specifically install Transfomers. However, a conda installation might be needed for ffmpeg[^1], which takes care of setting up PATH variables.
1. Install Python 3.10 or later. You can download it from [python.org](https://www.python.org/downloads/). During installation, **check "Add Python to PATH"**. No administrator rights are needed if you install for your user only.
From the Anaconda Prompt (which should now be installed in your system, find it with the search function), type or copy the following:
2. Run the installer. Open a terminal (Command Prompt on Windows, Terminal on Mac/Linux) in the project folder and run:
```
conda install -c conda-forge ffmpeg-python
python install.py
```
You can also choose not to use Anaconda (or miniconda), and use Python. In that case, you need to [download and install FFMPEG](https://ffmpeg.org/download.html) (and potentially add it to your PATH). See here for [WikiHow instructions](https://www.wikihow.com/Install-FFmpeg-on-Windows)
This will:
- Install all required packages (including bundled FFmpeg — no separate install needed)
- **Auto-detect your NVIDIA GPU** and ask if you want GPU acceleration
- No conda, no admin rights required
3. The main functionality comes from openai-whisper. See their [page](https://github.com/openai/whisper) for details. It also uses some additional packages (colorama, and customtkinter), install them with the following command:
```
pip install -r requirements.txt
```
4. Run the app:
1. For **Windows**: In the same folder as the *app.py* file, run the app from Anaconda prompt by running
```python app.py```
or with the batch file called run_Windows.bat (for Windows users), which assumes you have conda installed and in the base environment (This is for simplicity, but users are usually adviced to create an environment, see [here](https://conda.io/projects/conda/en/latest/user-guide/tasks/manage-environments.html#creating-an-environment-with-commands) for more info) just make sure you have the correct environment (right click on the file and press edit to make any changes).
3. For **Mac**: Haven't figured out a better way to do this, see [the instructions here](Mac_instructions.md)
Alternatively, you can install manually with `pip install -r requirements.txt`.
**Note** 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.
3. Run the app:
1. For **Windows**: double-click `run_Windows.bat` (it will auto-install on first run) or run:
```
python app.py
```
2. For **Mac / Linux**: run `./run_Mac.sh` (auto-installs on first run). See [Mac instructions](Mac_instructions.md) for details.
**Note** The first run with a given model will download it (~75 MB for base, ~500 MB for medium). After that, everything works offline.
## GPU Support
This program **does support running on NVIDIA GPUs**, which can significantly speed up transcription times. To use GPU acceleration, you need to have the correct version of PyTorch installed with CUDA support.
This program **does support running on NVIDIA GPUs**, which can significantly speed up transcription times. faster-whisper uses CTranslate2, which requires NVIDIA CUDA libraries for GPU acceleration.
### Installing PyTorch with CUDA Support
If you have an NVIDIA GPU and want to take advantage of GPU acceleration, you can install a CUDA-enabled version of PyTorch using:
### Automatic Detection
The `install.py` script **automatically detects NVIDIA GPUs** and will ask if you want to install GPU support. If you skipped it during installation, you can add it anytime:
```
pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121
pip install nvidia-cublas-cu12 nvidia-cudnn-cu12
```
**Note:** The command above installs PyTorch with CUDA 12.1 support. Make sure your NVIDIA GPU drivers are compatible with CUDA 12.1. You can check your CUDA version by running `nvidia-smi` in your terminal.
If you need a different CUDA version, visit the [PyTorch installation page](https://pytorch.org/get-started/locally/) to generate the appropriate installation command for your system.
**Note:** Make sure your NVIDIA GPU drivers are up to date. You can check by running `nvidia-smi` in your terminal. The program will automatically detect and use your GPU if available, otherwise it falls back to CPU.
### Verifying GPU Support
After installation, you can verify that PyTorch can detect your GPU by running:
After installation, you can verify that your GPU is available by running:
```python
import torch
print(torch.cuda.is_available()) # Should print True if GPU is available
print(torch.cuda.get_device_name(0)) # Should print your GPU name
import ctranslate2
print(ctranslate2.get_supported_compute_types("cuda"))
```
If GPU is not detected, the program will automatically fall back to CPU processing, though this will be slower.
If this returns a list containing `"float16"`, GPU acceleration is working.
## Usage
1. When launched, the app will also open a terminal that shows some additional information.
1. Launch the app — the built-in console panel at the bottom shows a welcome message and all progress updates.
2. Select the folder containing the audio or video files you want to transcribe by clicking the "Browse" button next to the "Folder" label. This will open a file dialog where you can navigate to the desired folder. Remember, you won't be choosing individual files but whole folders!
3. Enter the desired language for the transcription in the "Language" field. You can either select a language or leave it blank to enable automatic language detection.
4. Choose the Whisper model to use for the transcription from the dropdown list next to the "Model" label.
5. Enable the verbose mode by checking the "Verbose" checkbox if you want to receive detailed information during the transcription process.
6. Click the "Transcribe" button to start the transcription. The button will be disabled during the process to prevent multiple transcriptions at once.
7. Monitor the progress of the transcription with the progress bar.
8. Once the transcription is completed, a message box will appear displaying the transcribed text. Click "OK" to close the message box.
9. You can run the application again or quit the application at any time by clicking the "Quit" button.
5. Click the "Transcribe" button to start the transcription. The button will be disabled during the process to prevent multiple transcriptions at once.
6. Monitor progress in the embedded console panel — it shows model loading, per-file progress, and segment timestamps in real time.
7. Once the transcription is completed, a message box will appear displaying the result. Click "OK" to close it.
8. You can run the application again or quit at any time by clicking the "Quit" button.
## Jupyter Notebook
Don't want fancy EXEs or GUIs? Use the function as is. See [example](example.ipynb) for an 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)

102
app.py
View File

@@ -1,3 +1,5 @@
import os
import sys
import tkinter as tk
from tkinter import ttk
from tkinter import filedialog
@@ -5,10 +7,44 @@ from tkinter import messagebox
from src._LocalTranscribe import transcribe, get_path
import customtkinter
import threading
from colorama import Back
import colorama
colorama.init(autoreset=True)
import os
# ── Helper: redirect stdout/stderr into a CTkTextbox ──────────────────────
import re
_ANSI_RE = re.compile(r'\x1b\[[0-9;]*m') # strip colour codes
class _ConsoleRedirector:
"""Redirects output exclusively to the in-app console panel."""
def __init__(self, text_widget):
self.widget = text_widget
def write(self, text):
clean = _ANSI_RE.sub('', text) # strip ANSI colours
if clean.strip() == '':
return
# Schedule UI update on the main thread
try:
self.widget.after(0, self._append, clean)
except Exception:
pass
def _append(self, text):
self.widget.configure(state='normal')
self.widget.insert('end', text + ('\n' if not text.endswith('\n') else ''))
self.widget.see('end')
self.widget.configure(state='disabled')
def flush(self):
pass
# HuggingFace model IDs for non-standard models
HF_MODEL_MAP = {
'KB Swedish (tiny)': 'KBLab/kb-whisper-tiny',
'KB Swedish (base)': 'KBLab/kb-whisper-base',
'KB Swedish (small)': 'KBLab/kb-whisper-small',
'KB Swedish (medium)': 'KBLab/kb-whisper-medium',
'KB Swedish (large)': 'KBLab/kb-whisper-large',
}
@@ -18,7 +54,6 @@ firstclick = True
class App:
def __init__(self, master):
print(Back.CYAN + "Welcome to Local Transcribe with Whisper!\U0001f600\nCheck back here to see some output from your transcriptions.\nDon't worry, they will also be saved on the computer!\U0001f64f")
self.master = master
# Change font
font = ('Roboto', 13, 'bold') # Change the font and size here
@@ -28,6 +63,7 @@ class App:
path_frame.pack(fill=tk.BOTH, padx=10, pady=10)
customtkinter.CTkLabel(path_frame, text="Folder:", font=font).pack(side=tk.LEFT, padx=5)
self.path_entry = customtkinter.CTkEntry(path_frame, width=50, font=font_b)
self.path_entry.insert(0, os.path.join(os.getcwd(), 'sample_audio'))
self.path_entry.pack(side=tk.LEFT, fill=tk.X, expand=True)
customtkinter.CTkButton(path_frame, text="Browse", command=self.browse, font=font).pack(side=tk.LEFT, padx=5)
# Language frame
@@ -47,8 +83,13 @@ class App:
self.language_entry.bind('<FocusIn>', on_entry_click)
self.language_entry.pack(side=tk.LEFT, fill=tk.X, expand=True)
# Model frame
models = ['base.en', 'base', 'small.en',
'small', 'medium.en', 'medium', 'large']
models = ['tiny', 'tiny.en', 'base', 'base.en',
'small', 'small.en', 'medium', 'medium.en',
'large-v2', 'large-v3',
'───────────────',
'KB Swedish (tiny)', 'KB Swedish (base)',
'KB Swedish (small)', 'KB Swedish (medium)',
'KB Swedish (large)']
model_frame = customtkinter.CTkFrame(master)
model_frame.pack(fill=tk.BOTH, padx=10, pady=10)
customtkinter.CTkLabel(model_frame, text="Model:", font=font).pack(side=tk.LEFT, padx=5)
@@ -56,13 +97,8 @@ class App:
self.model_combobox = customtkinter.CTkComboBox(
model_frame, width=50, state="readonly",
values=models, font=font_b)
self.model_combobox.set(models[1]) # Set the default value
self.model_combobox.set('medium') # Set the default value
self.model_combobox.pack(side=tk.LEFT, fill=tk.X, expand=True)
# Verbose frame
verbose_frame = customtkinter.CTkFrame(master)
verbose_frame.pack(fill=tk.BOTH, padx=10, pady=10)
self.verbose_var = tk.BooleanVar()
customtkinter.CTkCheckBox(verbose_frame, text="Output transcription to terminal", variable=self.verbose_var, font=font).pack(side=tk.LEFT, padx=5)
# Progress Bar
self.progress_bar = ttk.Progressbar(master, length=200, mode='indeterminate')
# Button actions frame
@@ -71,6 +107,23 @@ class App:
self.transcribe_button = customtkinter.CTkButton(button_frame, text="Transcribe", command=self.start_transcription, font=font)
self.transcribe_button.pack(side=tk.LEFT, padx=5, pady=10, fill=tk.X, expand=True)
customtkinter.CTkButton(button_frame, text="Quit", command=master.quit, font=font).pack(side=tk.RIGHT, padx=5, pady=10, fill=tk.X, expand=True)
# ── Embedded console / log panel ──────────────────────────────────
log_label = customtkinter.CTkLabel(master, text="Console output", font=font, anchor='w')
log_label.pack(fill=tk.X, padx=12, pady=(8, 0))
self.log_box = customtkinter.CTkTextbox(master, height=220, font=('Consolas', 14),
wrap='word', state='disabled',
fg_color='#1e1e1e', text_color='#e0e0e0')
self.log_box.pack(fill=tk.BOTH, expand=True, padx=10, pady=(2, 10))
# Redirect stdout & stderr into the log panel (no backend console)
sys.stdout = _ConsoleRedirector(self.log_box)
sys.stderr = _ConsoleRedirector(self.log_box)
# Welcome message (shown after redirect so it appears in the panel)
print("Welcome to Local Transcribe with Whisper! \U0001f600")
print("Transcriptions will be saved automatically.")
print("─" * 46)
# Helper functions
# Browsing
def browse(self):
@@ -87,12 +140,22 @@ class App:
# Threading
def transcribe_thread(self):
path = self.path_entry.get()
model = self.model_combobox.get()
model_display = self.model_combobox.get()
# Ignore the visual separator
if model_display.startswith('─'):
messagebox.showinfo("Invalid selection", "Please select a model, not the separator line.")
self.transcribe_button.configure(state=tk.NORMAL)
return
model = HF_MODEL_MAP.get(model_display, model_display)
language = self.language_entry.get()
# Auto-set Swedish for KB models
is_kb_model = model_display.startswith('KB Swedish')
# Check if the language field has the default text or is empty
if language == self.default_language_text or not language.strip():
if is_kb_model:
language = 'sv'
elif language == self.default_language_text or not language.strip():
language = None # This is the same as passing nothing
verbose = self.verbose_var.get()
verbose = True # always show transcription progress in the console panel
# Show progress bar
self.progress_bar.pack(fill=tk.X, padx=5, pady=5)
self.progress_bar.start()
@@ -122,9 +185,10 @@ if __name__ == "__main__":
# Setting custom themes
root = customtkinter.CTk()
root.title("Local Transcribe with Whisper")
# Geometry
width,height = 450,275
root.geometry('{}x{}'.format(width,height))
# Geometry — taller to accommodate the embedded console panel
width, height = 550, 560
root.geometry('{}x{}'.format(width, height))
root.minsize(450, 480)
# Icon
root.iconbitmap('images/icon.ico')
# Run

4
build_setup.py
View File

@@ -1,7 +1,7 @@
from cx_Freeze import setup, Executable
build_exe_options = {
"packages": ['whisper','tkinter','customtkinter']
"packages": ['faster_whisper','tkinter','customtkinter']
}
executables = (
[
@@ -13,7 +13,7 @@ executables = (
)
setup(
name="Local Transcribe with Whisper",
version="1.2",
version="2.0",
author="Kristofer Rolf Söderström",
options={"build_exe":build_exe_options},
executables=executables

128
install.py Normal file
View File

@@ -0,0 +1,128 @@
"""
Installer script for Local Transcribe with Whisper.
Detects NVIDIA GPU and offers to install GPU acceleration support.
Usage:
python install.py
"""
import os
import subprocess
import sys
import shutil
import site
def detect_nvidia_gpu():
"""Check if an NVIDIA GPU is present."""
candidates = [
shutil.which("nvidia-smi"),
r"C:\Windows\System32\nvidia-smi.exe",
r"C:\Program Files\NVIDIA Corporation\NVSMI\nvidia-smi.exe",
]
for path in candidates:
if not path or not os.path.isfile(path):
continue
try:
r = subprocess.run(
[path, "--query-gpu=name", "--format=csv,noheader"],
capture_output=True, text=True, timeout=10,
)
if r.returncode == 0 and r.stdout.strip():
return True, r.stdout.strip().split("\n")[0]
except Exception:
continue
return False, None
def pip_install(*packages):
cmd = [sys.executable, "-m", "pip", "install"] + list(packages)
print(f"\n> {' '.join(cmd)}\n")
subprocess.check_call(cmd)
def get_site_packages():
for p in site.getsitepackages():
if p.endswith("site-packages"):
return p
return site.getsitepackages()[0]
def create_nvidia_pth():
"""Create a .pth startup hook that registers NVIDIA DLL directories."""
sp = get_site_packages()
pth_path = os.path.join(sp, "nvidia_cuda_path.pth")
# This one-liner runs at Python startup, before any user code.
pth_content = (
"import os, glob as g; "
"any(os.add_dll_directory(d) or os.environ.__setitem__('PATH', d + os.pathsep + os.environ.get('PATH','')) "
"for d in g.glob(os.path.join(r'" + sp.replace("'", "\\'") + "', 'nvidia', '*', 'bin')) "
"+ g.glob(os.path.join(r'" + sp.replace("'", "\\'") + "', 'nvidia', '*', 'lib')) "
"if os.path.isdir(d)) if os.name == 'nt' else None\n"
)
with open(pth_path, "w") as f:
f.write(pth_content)
print(f" Created CUDA startup hook: {pth_path}")
def verify_cuda():
"""Verify CUDA works in a fresh subprocess."""
try:
r = subprocess.run(
[sys.executable, "-c",
"import ctranslate2; "
"print('float16' in ctranslate2.get_supported_compute_types('cuda'))"],
capture_output=True, text=True, timeout=30,
)
return r.stdout.strip() == "True"
except Exception:
return False
def main():
print("=" * 55)
print(" Local Transcribe with Whisper — Installer")
print("=" * 55)
# Step 1: Base packages
print("\n[1/2] Installing base requirements...")
pip_install("-r", "requirements.txt")
print("\n Base requirements installed!")
# Step 2: GPU
print("\n[2/2] Checking for NVIDIA GPU...")
has_gpu, gpu_name = detect_nvidia_gpu()
if has_gpu:
print(f"\n NVIDIA GPU detected: {gpu_name}")
print(" GPU acceleration can make transcription 2-5x faster.")
print(" This will install ~300 MB of additional CUDA libraries.\n")
while True:
answer = input(" Install GPU support? [Y/n]: ").strip().lower()
if answer in ("", "y", "yes"):
print("\n Installing CUDA libraries...")
pip_install("nvidia-cublas-cu12", "nvidia-cudnn-cu12")
create_nvidia_pth()
print("\n Verifying CUDA...")
if verify_cuda():
print(" GPU support verified and working!")
else:
print(" WARNING: CUDA installed but not detected.")
print(" Update your NVIDIA drivers and try again.")
break
elif answer in ("n", "no"):
print("\n Skipping GPU. Re-run install.py to add it later.")
break
else:
print(" Please enter Y or N.")
else:
print("\n No NVIDIA GPU detected — using CPU mode.")
print("\n" + "=" * 55)
print(" Done! Run the app with: python app.py")
print("=" * 55)
if __name__ == "__main__":
main()

3
requirements.txt
View File

@@ -1,3 +1,2 @@
openai-whisper
faster-whisper
customtkinter
colorama

29
run_Mac.sh Normal file
View File

@@ -0,0 +1,29 @@
#!/bin/bash
# ============================================================
# Local Transcribe with Whisper — macOS / Linux launcher
# ============================================================
# Double-click this file or run: ./run_Mac.sh
# On first run it creates a venv and installs dependencies.
# ============================================================
set -e
cd "$(dirname "$0")"
# Create .venv if it doesn't exist
if [ ! -f ".venv/bin/python" ]; then
echo "Creating virtual environment..."
python3 -m venv .venv
fi
PYTHON=".venv/bin/python"
# Install dependencies on first run
if ! "$PYTHON" -c "import faster_whisper" 2>/dev/null; then
echo "First run detected — running installer..."
"$PYTHON" install.py
echo
fi
echo "Starting Local Transcribe..."
"$PYTHON" app.py

26
run_Windows.bat
View File

@@ -1,5 +1,23 @@
@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 app.py
REM Create .venv on first run if it doesn't exist
if not exist ".venv\Scripts\python.exe" (
echo Creating virtual environment...
python -m venv .venv
if errorlevel 1 (
echo ERROR: Failed to create virtual environment. Is Python installed and on PATH?
pause
exit /b 1
)
)
set PYTHON=.venv\Scripts\python.exe
REM Check if dependencies are installed
%PYTHON% -c "import faster_whisper" 2>nul
if errorlevel 1 (
echo First run detected - running installer...
%PYTHON% install.py
echo.
)
echo Starting Local Transcribe...
%PYTHON% app.py

2
sample_audio/transcriptions/Armstrong_Small_Step.txt
View File

@@ -1,2 +1,2 @@
Armstrong_Small_Step
[0:00:00 --> 0:00:29.360000]: alumnfeldaguyrjarna om det nya skirprå kızım om det där föddarna hatt splittar, do nackrott,
[0:00:00 --> 0:00:07]: That's one small step for man, one giant leap for mankind.

4
sample_audio/transcriptions/Axel_Pettersson_röstinspelning.txt
View File

@@ -1,4 +1,2 @@
Axel_Pettersson_röstinspelning
[0:00:00 --> 0:00:06]: Hej, jag heter Raxel Patterson, jag får att se över UR 1976.
[0:00:06 --> 0:00:12.540000]: Jag har varit Wikipedia-périonsen 2018 och jag har översat röst-intro-
[0:00:12.540000 --> 0:00:15.540000]:-projektet till svenska.
[0:00:00 --> 0:00:15]: Hej, jag heter Axel Pettersson, jag föddes i Örebro 1976. Jag har varit Wikipedia sen 2008 och jag har översatt röstintroduktionsprojektet till svenska.

153
src/_LocalTranscribe.py
View File

@@ -1,11 +1,56 @@
import os
import sys
import datetime
import site
from glob import glob
import whisper
from torch import backends, cuda, Generator
import colorama
from colorama import Back,Fore
colorama.init(autoreset=True)
# ---------------------------------------------------------------------------
# CUDA setup — must happen before importing faster_whisper / ctranslate2
# ---------------------------------------------------------------------------
def _setup_cuda_dlls():
"""Add NVIDIA pip-package DLL dirs to the DLL search path (Windows only).
pip-installed nvidia-cublas-cu12 / nvidia-cudnn-cu12 place their .dll
files inside the site-packages tree. Python 3.8+ on Windows does NOT
search PATH for DLLs loaded via ctypes/LoadLibrary, so we must
explicitly register every nvidia/*/bin and nvidia/*/lib directory using
os.add_dll_directory *and* prepend them to PATH (some native extensions
still rely on PATH).
"""
if sys.platform != "win32":
return
try:
for sp in site.getsitepackages():
nvidia_root = os.path.join(sp, "nvidia")
if not os.path.isdir(nvidia_root):
continue
for pkg in os.listdir(nvidia_root):
for sub in ("bin", "lib"):
d = os.path.join(nvidia_root, pkg, sub)
if os.path.isdir(d):
os.environ["PATH"] = d + os.pathsep + os.environ.get("PATH", "")
try:
os.add_dll_directory(d)
except (OSError, AttributeError):
pass
except Exception:
pass
_setup_cuda_dlls()
from faster_whisper import WhisperModel
def _detect_device():
"""Return (device, compute_type) for the best available backend."""
try:
import ctranslate2
cuda_types = ctranslate2.get_supported_compute_types("cuda")
if "float16" in cuda_types:
return "cuda", "float16"
except Exception:
pass
return "cpu", "int8"
# Get the path
@@ -16,12 +61,12 @@ def get_path(path):
# Main function
def transcribe(path, glob_file, model=None, language=None, verbose=False):
"""
Transcribes audio files in a specified folder using OpenAI's Whisper model.
Transcribes audio files in a specified folder using faster-whisper (CTranslate2).
Args:
path (str): Path to the folder containing the audio files.
glob_file (list): List of audio file paths to transcribe.
model (str, optional): Name of the Whisper model to use for transcription.
model (str, optional): Name of the Whisper model size to use for transcription.
Defaults to None, which uses the default model.
language (str, optional): Language code for transcription. Defaults to None,
which enables automatic language detection.
@@ -38,59 +83,67 @@ def transcribe(path, glob_file, model=None, language=None, verbose=False):
- The function downloads the specified model if not available locally.
- The transcribed text files will be saved in a "transcriptions" folder
within the specified path.
- Uses CTranslate2 for up to 4x faster inference compared to openai-whisper.
- FFmpeg is bundled via the PyAV dependency — no separate installation needed.
"""
# Check for GPU acceleration and set device
if backends.mps.is_available():
device = 'mps'
Generator('mps').manual_seed(42)
elif cuda.is_available():
device = 'cuda'
Generator('cuda').manual_seed(42)
else:
device = 'cpu'
Generator().manual_seed(42)
SEP = "─" * 46
# Load model on the correct device
model = whisper.load_model(model, device=device)
# Start main loop
files_transcripted=[]
# ── Step 1: Detect hardware ──────────────────────────────────────
device, compute_type = _detect_device()
print(f"⚙ Device: {device} | Compute: {compute_type}")
# ── Step 2: Load model ───────────────────────────────────────────
print(f"⏳ Loading model '{model}' — downloading if needed...")
whisper_model = WhisperModel(model, device=device, compute_type=compute_type)
print("✅ Model ready!")
print(SEP)
# ── Step 3: Transcribe files ─────────────────────────────────────
total_files = len(glob_file)
print(f"📂 Found {total_files} item(s) in folder")
print(SEP)
files_transcripted = []
file_num = 0
for file in glob_file:
title = os.path.basename(file).split('.')[0]
print(Back.CYAN + '\nTrying to transcribe file named: {}\U0001f550'.format(title))
file_num += 1
print(f"\n{'─' * 46}")
print(f"📄 File {file_num}/{total_files}: {title}")
try:
result = model.transcribe(
segments, info = whisper_model.transcribe(
file,
language=language,
verbose=verbose
beam_size=5
)
files_transcripted.append(result)
# Make folder if missing
try:
os.makedirs('{}/transcriptions'.format(path), exist_ok=True)
except FileExistsError:
pass
# Create segments for text files
start = []
end = []
text = []
for segment in result['segments']:
start.append(str(datetime.timedelta(seconds=segment['start'])))
end.append(str(datetime.timedelta(seconds=segment['end'])))
text.append(segment['text'])
# Save files to transcriptions folder
with open("{}/transcriptions/{}.txt".format(path, title), 'w', encoding='utf-8') as file:
file.write(title)
for i in range(len(result['segments'])):
file.write('\n[{} --> {}]:{}'.format(start[i], end[i], text[i]))
# Skip invalid files
except RuntimeError:
print(Fore.RED + 'Not a valid file, skipping.')
pass
# Check if any files were processed.
if len(files_transcripted) > 0:
output_text = 'Finished transcription, {} files can be found in {}/transcriptions'.format(len(files_transcripted), path)
# Stream segments as they are decoded
segment_list = []
with open("{}/transcriptions/{}.txt".format(path, title), 'w', encoding='utf-8') as f:
f.write(title)
for seg in segments:
start_ts = str(datetime.timedelta(seconds=seg.start))
end_ts = str(datetime.timedelta(seconds=seg.end))
f.write('\n[{} --> {}]:{}'.format(start_ts, end_ts, seg.text))
f.flush()
if verbose:
print(" [%.2fs → %.2fs] %s" % (seg.start, seg.end, seg.text))
else:
output_text = 'No files elligible for transcription, try adding audio or video files to this folder or choose another folder!'
# Return output text
print(" Transcribed up to %.0fs..." % seg.end, end='\r')
segment_list.append(seg)
print(f"✅ Done — saved to transcriptions/{title}.txt")
files_transcripted.append(segment_list)
except Exception:
print('⚠ Not a valid audio/video file, skipping.')
# ── Summary ──────────────────────────────────────────────────────
print(f"\n{SEP}")
if len(files_transcripted) > 0:
output_text = f"✅ Finished! {len(files_transcripted)} file(s) transcribed.\n Saved in: {path}/transcriptions"
else:
output_text = '⚠ No files eligible for transcription — try another folder.'
print(output_text)
print(SEP)
return output_text