AI Audio Enhancer
car-diagnosis (cardiag)
An open-source ML pipeline that diagnoses car faults from audio by cleaning sound, embedding with CLAP, and training calibrated linear classifiers.
car-diagnosis (cardiag)
What is car-diagnosis (cardiag)?
cardiag is an end-to-end audio-ML pipeline that scrapes fault-sound clips from YouTube/TikTok, cleans audio to isolate mechanical sound, embeds it with a frozen CLAP model, and trains small linear heads to triage the fault. It exposes a CLI and a live web app for inference.
car-diagnosis (cardiag) vs Similar AI Tools
| Pricing Model | Free | Free, Freemium | Free | Free, Free Trial, Paid |
| Free Credits | ||||
| Key Features |
|
|
|
|
| Pros |
|
|
|
|
| Cons |
|
|
|
|
| Best For |
|
|
|
|
How to use car-diagnosis (cardiag)?
- 1Clone the repository: git clone <repo> && cd car-diagnosis
- 2Set up a Python 3.11 environment with dependencies: uv venv && source .venv/bin/activate && uv pip install -e ".[scrape,web,dev,viz]"
- 3Run a verification training: cardiag train --fixtures
- 4Diagnose a clip: cardiag diagnose <clip.wav> (or use the web app via cardiag serve --model models)
car-diagnosis (cardiag) Key Features
- Scrapes audio clips from YouTube and TikTok
- Cleans audio to isolate mechanical sound from speech, music, and noise
- Uses a frozen CLAP model for audio embedding
- Trains small linear heads for calibrated triage (fault/normal, zone, part)
- Returns calibrated uncertainty (ECE ≈ 0.04) and 'UNCERTAIN' when appropriate
- Pre-trained model ships with the repository
- CLI and web app interfaces
car-diagnosis (cardiag) Use Cases
- Quick preliminary diagnosis of car noises from a phone recording
- Educational demo of audio-ML pipeline design
- Triage aid for mechanics to narrow down possible fault locations
- Dataset building and custom training for specific car models
car-diagnosis (cardiag) Pricing & Free Credits
car-diagnosis (cardiag) currently operates on a Free model.
This tool is completely free to use
car-diagnosis (cardiag) Pros & Cons
Pros
- Honest calibration with uncertainty awareness
- Reusable pipeline for other audio classification tasks
- Includes pre-trained model and synthetic demo clip for immediate use
Cons
- Modest accuracy on crude phone audio (AUROC 0.79 for fault/normal)
- Not safety-critical or a substitute for professional diagnostics
- Requires Python environment setup and some technical knowledge
What is car-diagnosis (cardiag) best for?
- Car owners wanting a preliminary, non-intrusive check of engine/mechanical sounds
- Developers and researchers exploring audio-ML pipelines
- Mechanics using audio as a triage tool before physical inspection