Transcripta Lite

Transcripta Lite is a modern desktop application that converts speech and audio files into accurate text using OpenAI’s Whisper model.
It provides an intuitive, clean interface built with CustomTkinter, offering users full control over model selection, CPU/GPU usage, and output formatting.

The app allows you to:

• Transcribe audio or video files directly to text.(multilingual)(+90 language)
• Record speech in real-time and save it in multiple formats (WAV, MP3, FLAC).
• Copy, edit, and save the transcribed text in .txt or .docx formats.
• Choose between light and dark themes.
• Instantly switch the entire UI language — supports English, Arabic, and up to 10 languages.
• Select from multiple Whisper models (Tiny, Base, Small) to balance between speed and accuracy.
• Control how many CPU threads are used for transcription to boost performance when running without a GPU.

Transcripta Lite is designed for creators, journalists, students, and professionals who need fast and reliable speech-to-text transcription on their desktop — without any complicated setup. Transcripta Lite is a free and open-source desktop application that runs 100% offline, ensuring complete privacy and security for your recordings.

---

Installation & Setup

1. Prerequisites

Before running Transcripta Lite , make sure you have:

• Python 3.10 or higher installed.
• FFmpeg available in your system PATH (required for audio conversion).
• A stable offline environment (the app runs 100% locally).

FFmpeg Setup (Required)

Transcripta Lite requires FFmpeg for audio conversion and playback.
If FFmpeg is not installed, please follow our detailed step-by-step guide here:

How to install FFmpeg on Windows, macOS, and Linux

• (Optional) NVIDIA GPU for faster transcription (CUDA supported)
• Check with: nvidia-smi or python -m torch.cuda.is_available

2. Install Required Packages

Depending on your device, you can choose one of the two requirement files:

For CPU users:

pip install -r requirements_cpu.txt

For GPU users (with NVIDIA GPU + CUDA installed):

pip install -r requirements_gpu.txt

Before installation, you can check whether your computer has an NVIDIA GPU (for CUDA support):

Follow this guide to check your GPU availability or simply run:

python Get_gpu_Cpu_Names.py

Install Microsoft Visual C++ Redistributable (Required for Windows)

Some dependencies used by Transcripta Lite — such as PyQt5, PyTorch, and certain audio modules — require the Microsoft Visual C++ Redistributable to run properly on Windows.

Download & install the latest supported version from Microsoft:

Download (Official Microsoft Page):

https://aka.ms/vs/17/release/vc_redist.x64.exe

After installing, restart your computer.

Check our installation guide step by step

3. After installing all dependencies, simply run:

run_transcripta.bat

The launcher will automatically start Transcripta Lite.

• The Transcripta Lite window will open automatically.

4. If you encounter missing package errors, install them manually using:

• pip install package_name

Example

• pip install customtkinter

---

How to Use Transcripta Lite

Once Transcripta Lite is installed and running, follow these simple steps to start transcribing your audio or video files:

1. Choose an Audio or Video File

• Click Browse in the left panel to select a file (.wav, .mp3, .mp4, .m4a, .ogg, .opus, .flac, etc.).
• The selected file path will appear in the interface, and a short notification will confirm it.

2. (Optional) Record Your Voice

• Click Record to capture live speech directly from your microphone.
• When finished, click Stop Record and choose your desired audio format (WAV, MP3, or FLAC).
• The recording will automatically be available for transcription.

3. Select the Model

• Choose between:
  • Tiny – fastest, less accurate
  • Base – balanced performance
  • Small – slower but highest accuracy
• This lets you control the trade-off between speed and accuracy depending on your needs.

4. Optimize CPU/GPU Usage

• If your computer has a GPU, the app will automatically detect and use it for faster transcription.
• If no GPU is detected:
  • You can enable Use full CPU to utilize all processor threads.
  • Or manually adjust the thread slider to balance speed and system load.

5. Start the Transcription

• Click Start to begin transcribing.
• The progress bar and ETA (Estimated Time Remaining) will update as Whisper processes your file.
• When finished, the transcribed text appears instantly in the editor area.

6. Review & Edit

• You can copy, edit, or clear text directly inside the built-in editor.
• For advanced editing, click Edit to open the text in a PyQt-based rich editor window.

7. Save Your Transcription

• Click Save to export your text as:
  • .txt (plain text)
  • .docx (Word document)
• All files are saved locally — no internet connection is ever required.

8. Additional Options

• Change the interface theme (Light or Dark).
• Instantly switch the UI language from the dropdown menu.
• View all messages, logs, and alerts in the Notifications panel on the right.

---

Tip:
If you close the app while a transcription is running, it will stop automatically to prevent corrupted files.
All settings (language, theme, and model selection) are remembered between sessions.

---

Features Overview

Transcripta Lite combines performance, simplicity, and flexibility — all in lightweight desktop package.
Here’s an overview of its key features:

Audio & Video Transcription

• Supports all major formats: WAV, MP3, MP4, M4A, FLAC, OGG, OPUS, and more.
• Converts speech to text using OpenAI’s Whisper model for top-tier accuracy.
• Works fully offline, keeping your audio and text private on your device.

Real-Time Recording

• Record directly from your microphone without external tools.
• Save recordings automatically in WAV, MP3, or FLAC format.
• Option to convert files after recording for compact storage.

Model Selection & Performance Control

• Choose between multiple Whisper models:
  • Tiny (Fastest)
  • Base (Balanced)
  • Small (Most Accurate)
• Adjust CPU thread usage to boost performance on non-GPU systems.
• Automatic GPU detection for NVIDIA CUDA and Apple Metal acceleration.

Smart Text Editor

• Built-in CustomTkinter editor with undo/redo and keyboard shortcuts.
• Integrated PyQt5 text window for larger or formatted editing.
• Supports copy, clear, and quick export functions.

Multi-Language Interface

• Instantly switch the app UI between multiple languages (English, Arabic, and up to 10 others).
• All translations are handled dynamically without restarting the app.

Modern User Interface

• Elegant and minimal design built with CustomTkinter.
• Supports Light and Dark modes.
• Resizable, responsive layout across all desktop systems.

Flexible Export Options

• Save transcripts as plain text (.txt) or Word document (.docx).
• Automatic creation of recording folders.
• Clear and timestamped notifications panel to track all app activity.

Privacy

• 100% offline — no internet required for transcription or recording.
• Your audio and text files never leave your device.
• Free and open-source under the Apache License.

---

Troubleshooting

If you encounter an issue, check the quick references below:

Issue	Possible Cause	Solution
App says “FFmpeg not found”	FFmpeg not installed or not added to PATH	Follow FFmpeg setup guide
“No GPU detected” message	Running on CPU only	Check GPU availability
“Whisper not installed”	Missing dependencies	Run pip install -r requirements_cpu.txt or requirements_gpu.txt
Audio won’t play	Missing player or unsupported format	Ensure FFmpeg or VLC is installed
Save as DOCX failed	python-docx not installed	Run pip install python-docx

Tip: All errors are also logged in the app’s notification panel with timestamps.