FreeHarmony
Measure your facial proportions with real geometry — private, free, and fully in your browser
AI: local or online
Where FreeHarmony’s AI runs, and what it costs
Prices are for one typical use: one request of about 1,500 words sent and 375 words back. A higher quality score is better.
On your computer
- Price
- Free per use. Your computer does the work.
- Runs with
- MediaPipe, Ollama
- Memory
- Llama 3.1 8B is a 4.9 GB download at 4-bit. gpt-oss 20B runs on systems with as little as 16GB memory. Faster with Apple silicon or a graphics card.
- Quality
- Llama 3.1 8B: 7
publik API
- Price
- $0.0100 per use, $10.00 per 1,000 usespublik-balanced, the default level
- Quality
- MiMo-V2.6-Pro: 46
- Setup
- Not built into FreeHarmony yet. A publik API key works wherever FreeHarmony takes an OpenAI-compatible address.
Your own key
- Price
- $0.0013 per use, $1.31 per 1,000 usesMiMo-V2.6-Pro at OpenRouter’s list price, before its fee for buying usage
- Quality
- MiMo-V2.6-Pro: 46
- Setup
- Open a provider account, add a card, paste the key into FreeHarmony.
Quality and price, side by side
| Model | Quality | Per 1,000 uses |
|---|---|---|
| On your computer (Ollama, 4-bit download size) | ||
| Granite 4.2 3B2.2 GB | 9 | $0 |
| Phi-4 Mini2.5 GB | 6 | $0 |
| Llama 3.1 8B4.9 GB | 7 | $0 |
| gpt-oss 20B14 GB | 9 | $0 |
| Gemma 4 31B20 GB | 19* | $0 |
| Qwen3.5 35B-A3B24 GB | 19* | $0 |
| publik API | ||
| publik-fastGLM-5.3 Flash | 42 | $1.00 |
| publik-balancedMiMo-V2.6-Pro · FreeHarmony | 46 | $10.00 |
| publik-smartGPT-6 Sol | 48 | $18.00 |
Quality: Artificial Analysis Intelligence Index v4.3.2, read 2026-09-22 (publik API models 2026-09-25); * = estimated by Artificial Analysis. Sizes: the Ollama library, read 2026-09-22.
What you pay for
- On your computer: nothing per use. You pay in disk space, memory and electricity, at lower quality.
- publik API: publik’s published price for each use, in dollars, from your publik balance. It is above the model’s cost; the difference runs publik and pays the app’s builder.
- Your own key: the provider’s price, billed to an account you keep with the provider.
How to install FreeHarmony
Every step written out. No terminal experience needed. Pick your setup.
README
Open in GitHubFreeHarmony
Free, open-source facial harmony analysis. Every metric. No paywall. Ever.
FreeHarmony measures your facial proportions — canthal tilt, midface ratio, jaw-to-cheekbone width, symmetry, and a dozen more — using real landmark geometry that runs entirely in your browser. It exists because the app it replaces charges $4.99/week to show you numbers your own webcam already computed, teases you with a flattering fake score, and then reveals a lower "real" one when you decline to pay.
We think that's garbage. So:
- Everything is free. Every metric, every sub-score, the full report, side profile, scan history. There is no Pro tier because there is no tier.
- The score is math, and only math. Deterministic geometry from MediaPipe face
landmarks. Same photo in, same score out — auditable in
packages/engine, which has no dependencies and 100% reproducible tests. - Your face never leaves your device. Landmarking and scoring are fully client-side. The only thing that can ever leave your browser: an AI second-opinion call to a provider you configure (your local Ollama, or your own Claude key). Off by default.
- No dark patterns. One honest score, shown once. If your photo is too blurry or turned too far, we tell you to retake it instead of guessing.
How scoring works
- MediaPipe Face Landmarker (WASM, in-browser) finds 478 facial landmarks.
packages/enginenormalizes them (de-mirror, pixel-space, roll-correction) and refuses photos it can't measure honestly (excessive yaw/pitch, blur, occlusion, closed eyes).- Seventeen metrics are computed from published facial-aesthetics conventions, each scored against an ideal band with a smooth falloff curve, then rolled up into four area scores and one Harmony %.
- Optionally, a vision model you control cross-examines the landmarks ("did the jaw point land on an ear?") and can flag a scan as unreliable — it can never change the score.
Scores are photo-based estimates for entertainment and self-improvement, not clinical measurements — lighting, lens distance, and pose all move the numbers. The app says this in-product, and the advice engine hard-refuses surgical/medication coaching and actively discourages dangerous community practices.
License
AGPL-3.0. Use it, fork it, ship it — but if you serve a modified version to users, you must publish your source. Paywalling a fork of the anti-paywall app should at least be embarrassing in public.