Terms of use

Who makes Mona

Mona is made by Liam DB, one person, not a company. There is no team, no customer service and no promise of a reply. These terms were written by the author himself, without a lawyer: they simply and honestly say what is allowed and what is not.

What they apply to

They apply to the site mona.devliam.com, to the local app published on GitHub and to the AI models that come with it. Using Mona means accepting these terms. If you don't agree, don't use it.

What you can do

• use Mona for free, for yourself, without an account;
• learn, experiment, check your own non-commercial projects;
• show your results, crediting Mona (see point 5).

The full license (“Mona Personal Use License”) is in the LICENSE file. For the code and the models, that is what counts.

What you can't do

• copy, publish, share or redistribute Mona, its models or a modified version;
• sell it, charge for access, or use it, directly or indirectly, to make money (work for clients, a paid service, a company's product, a site funded by advertising);
• offer it online as a service to other people, even for free;
• use its models to train or improve other AI models;
• remove the copyright notices or the requested credit;
• overload, disrupt or get around how the site works, or use it for anything illegal.

For any other use, ask the author first, in writing. The answer may be yes.

Credit

If you show or publish something that uses Mona or its models (project, article, video, repository, web page), credit it visibly: “Mona by Liam DB, https://github.com/LiamLitle/mona-local”.

What Mona can do, and its limits

Gaze maps, brain responses and scores are estimates made by AI models. They were measured on cases the models had never seen, and all the numbers are published on the benchmark page, including the bad ones:

• the predicted gaze looks like people's real gaze, without being identical (correlation 0.68 on UEyes, 0.56 on web pages from another dataset);
• the detector finds text well, icons (F1 0.36) and form fields (F1 0.49) less well;
• a score calculated from a plain image differs on average by 5.5 points out of 100 from the score calculated from the page code;
• the brain model was trained on data from a single person and has not been checked on web pages: it's an indication, not a measurement.

Mona measures readability and attention, not taste. It replaces neither a test with real people nor an accessibility audit, and proves no compliance (WCAG, RGAA or other).

What works today

On the site, you can drop an image or share a tab from your browser. Analyzing an address (URL) directly on the site is not available yet: it needs a server that isn't online for now. The local app, however, can already analyze an address.

Your images and the sites you analyze

You remain responsible for what you analyze. Only analyze sites and images you have the right to view and use, and respect their terms and their authors' rights. With the local app, it's your computer that visits the sites you enter.

Availability

Mona is provided “as is”, without warranty. It's a personal project: it can change, stop or disappear at any time, without notice.

Liability

To the extent permitted by law, the author is not liable for direct or indirect damage related to the use of Mona, or for decisions you make based on its results. Nothing here removes a liability that cannot legally be removed, or your consumer rights.

Training data and other people's licenses

The models were trained with datasets and pretrained models that have their own terms (see the NOTICE file). In particular, the brain model comes from THINGS images reserved for research: it may only be used for non-commercial research. Mona doesn't give you more rights than those terms allow.

Changes and governing law

These terms may change. The online version is the one that counts, and the date at the top of the page changes with every update. They are governed by the law of the country where the author lives, without taking away the rules that protect you as a consumer.