This repository provides a Google Colab-based implementation of OpenAI’s Whisper AI for transcribing audio files into text.
Simply clone the repo, run the notebook, upload an audio file, and get an accurate transcription in seconds!
✅ No setup required – Just open the .ipynb file in Colab
✅ User-friendly – Upload an audio file, run a single command, and get transcriptions
✅ Supports multiple languages with the Whisper medium model
✅ Works with multiple audio formats including MP3, WAV, M4A, and FLAC
Whisper AI supports a variety of common audio formats. You can upload files in any of the following formats:
| Format | File Extension | Description |
|---|---|---|
| MP3 | .mp3 |
Compressed audio format, widely used |
| WAV | .wav |
High-quality, uncompressed audio |
| M4A | .m4a |
Common format for iOS recordings |
| FLAC | .flac |
Lossless audio format, high fidelity |
💡 Note: Files in other formats can be converted using FFmpeg before transcription.
Open a new Colab notebook and run the following:
!git clone https://github.com/asadsandhu/whisper-audio-to-text.git
%cd whisper-audio-to-textRun the following commands one by one to set up Whisper AI and FFmpeg:
# Install Whisper AI
!pip install git+https://github.com/openai/whisper.git# Install FFmpeg (required for audio processing)
!sudo apt update && sudo apt install ffmpegIf you want to transcribe your own audio file, run this in a Colab cell:
from google.colab import files
uploaded = files.upload()🔹 After uploading your file, remember its name.
🔹 In step 5, replace "Sample.mp3" with the name of your uploaded file.
If you don’t want to upload a file and just want to test Whisper AI, run this command:
!wget https://github.com/asadsandhu/whisper-audio-to-text/raw/main/Sample.mp3 -O Sample.mp3🔹 This will automatically download a sample MP3 file (Sample.mp3) from GitHub.
🔹 If you already uploaded a file, skip this step.
Now, transcribe the audio file with Whisper:
!whisper "Sample.mp3" --model medium🔹 If you uploaded your own file in step 3, replace "Sample.mp3" with your file name.
🔹 You can change the model by replacing medium with any of the available Whisper models.
| Model | Size | Speed | Accuracy | Best For |
|---|---|---|---|---|
tiny |
39 MB | 🚀 Very Fast | ❌ Low Accuracy | Quick testing, low-end devices |
base |
74 MB | ⚡ Fast | 🔸 Moderate Accuracy | Short recordings, general use |
small |
244 MB | ⚡ Moderate | 🔹 Good Accuracy | Standard transcription tasks |
medium |
769 MB | ⏳ Slower | ✅ High Accuracy | Most users, multilingual support |
large |
1550 MB | 🕒 Slowest | 🔥 Best Accuracy | Research, high-quality needs |
🔹 Larger models provide better accuracy but take longer to process.
🔹 If speed is a priority, use small or base. If accuracy is more important, use medium or large.
🔹 Example: Using the large model
!whisper "Sample.mp3" --model large| Scenario | Steps to Follow |
|---|---|
| Upload your own audio file | Run Step 3 → Skip Step 4 → Run Step 5 (rename file) |
| Use the sample audio file | Skip Step 3 → Run Step 4 → Run Step 5 |
A sample MP3 file (Sample.mp3) is included in this repository for quick testing. You can use it to check the functionality before uploading your own files.
- Google Colab (Runs in the cloud, no local installation needed)
- Python 3.7+
- Whisper AI (
pip install whisper) - FFmpeg (pre-installed in Colab)
✔ Minimal Effort – Just upload and run, no extra configurations
✔ Highly Accurate – Uses Whisper’s medium model for precise transcriptions
✔ Cloud-Based – Runs on Google Colab’s free GPU resources
✔ Multiple Formats – Supports MP3, WAV, M4A, FLAC
Contributions are welcome! If you find any issues or have ideas for improvements, feel free to submit a pull request or open an issue.
This project is licensed under the MIT License.
If you find this repository helpful, feel free to star 🌟 it on GitHub! 🚀