ShowMeme

Este documento está em inglês. As versões em inglês e polonês são as que valem.

AI Notice

Last updated: 2026-11-05

ShowMeme uses artificial intelligence to edit photos. This page lists the models we use, explains how we mark AI results and what AI gets wrong. It is our transparency notice under art. 50 of the EU AI Act (Regulation (EU) 2024/1689).

1. Every result is made by AI

Every image you get from the studio was created or edited by an AI model.

When you pick an effect, an AI model redraws your photo following an instruction (prompt) we wrote for that effect. You cannot type your own prompt. Only two of your answers on the photo page change the wording of the instruction: "How should I draw you?" and "Who is in the photo?" (woman, man or skipped). Each further effect starts from the previous result, so the changes add up.

The AI does not recognise who you are, check your identity or profile you. The only decision software makes on its own is the moderation block described in the Content Rules.

2. How we mark results

Each result carries a visible watermark and a label inside the file.

  • Visible watermark: The Shomi mark and "showmeme.app" in the bottom-right corner of every result, including the file you download. Link previews of shared results show the watermarked image too.
  • Machine-readable label: We write into the file's metadata the IPTC Digital Source Type value "trainedAlgorithmicMedia", which tells software that a trained AI model created the image, plus the description "Edited with generative AI by ShowMeme (showmeme.app)" and ShowMeme as the creating tool. Link preview images carry the same label.

Some apps and social networks strip metadata when you upload an image, and cropping removes the watermark. That is why our Terms of Service forbid removing either. We do not yet add signed content credentials (C2PA) or an invisible watermark.

3. Models we use

deAPI runs the model that edits your photo, and a separate model checks images for blocked content.

  • Image editing: FLUX.2 klein 4B by Black Forest Labs, run for us by deAPI (CoinAxe Ltd, Malta). If we switch to another model, we will update this page.
  • Moderation: Gemma 3 12B by Google, with Gemma 3 27B as a backup, reached through OpenRouter and run only by DeepInfra, which keeps no copy of the image.

We do not use your photos or results to train AI models, and deAPI's privacy policy says it does not use customer content for training either. The Privacy Policy explains where your photo goes and for how long.

4. What AI gets wrong

AI results can surprise you, so check them before you share.

  • The result may look less like you, or like someone else.
  • AI may change your skin tone, age, face shape, body or hair more than the effect promises.
  • It can add odd details, such as extra fingers, garbled text or a broken background.
  • The same effect gives different people different results, and two people may get similar ones.
  • Models learn from large collections of images and can repeat stereotypes about sex, age or origin.
  • Horror effects add fake blood and wounds on purpose.
  • Moderation can block a harmless photo or miss a bad one.

If a result is offensive or harmful, delete it with "Delete now" and tell us at [email protected], so we can fix the effect.

5. Sharing AI results

Keep the watermark, and do not pass a result off as a real photo.

  • Keep the watermark and the metadata.
  • Do not present a result as an unedited photo or as a real event.
  • Use the "AI-generated" label wherever a platform offers one.
  • Edit other people only with their consent, and never make deepfakes of public figures (Content Rules).

6. How to report misuse

Report a result with the Report button or by e-mail.

If you see a ShowMeme result used to deceive, harass or impersonate someone, tap "Report" on the result page or write to [email protected] with the link. A person reviews every report. The Content Rules explain what happens next.

7. Legal background

The AI Act requires AI images to be marked, and we mark ours from the first result.

Art. 50 of the AI Act applies from 2 August 2026. It requires providers of AI systems that generate images to mark the output in a machine-readable way (art. 50(2)), and requires deployers who publish deepfakes, outside personal non-professional use, to disclose that the content was generated or manipulated by AI (art. 50(4)). Under the European Commission's guidelines of July 2026, we are a provider under art. 50(2), because we offer the image generator under our own name. We mark every result visibly and with IPTC metadata from the first day the studio runs. We plan to add signed content credentials (C2PA) and an invisible watermark, following the Commission's code of practice on marking of 10 June 2026. When you use ShowMeme for purely personal, non-professional purposes, the AI Act's deployer obligations do not apply to you (art. 2(10)), but our Terms still require you to keep the marks when you share.