Talk:Text-to-image model

Todos

• Example caption-image pairs.
  • Look at commons:Category:Artificial intelligence art, and see which cited papers make their figures available under an appropriate license.
    • Can't find freely licensed examples from Mansimov, but if we e-mailed him I bet there's a good chance he would slap a CC license on http://www.cs.toronto.edu/~emansim/cap2im.html or something. Actually, looks like there's consensus on Commons to treat AI-generated images as non-copyrightable, so that makes collecting illustrations a lot easier.
• Expand history post-2016
• More sources for architectures section
• "Evaluation" section
• Section listing notable models (which might eventually be appropriate to split off into a separate SA list article)
• Add inbound links (may also want to think about slightly rescoping Artificial intelligence art for a cleaner separation of concerns)

Colin M (talk) 16:54, 7 September 2022 (UTC)

This article has a good lead.

The first sentence gives a clear and concise description of the topic. The first paragraph contains the most pertinent information about the topic without going into unnecessary detail. All information introduced in the lead is relevant and later expanded on in the article.

Muh tee us? (talk) 00:49, 9 September 2022 (UTC)

Wiki Education assignment: Technology and Culture

This article was the subject of a Wiki Education Foundation-supported course assignment, between 21 August 2023 and 15 December 2023. Further details are available on the course page. Student editor(s): Nadpnw (article contribs).

— Assignment last updated by Thecanyon (talk) 05:32, 12 December 2023 (UTC)

Edit request

• Text-to-image model (edit | talk | history | links | watch | logs)

• Page: Text-to-image model
• Requested section: Quality evaluation
• Reason for posting here: I cannot add a topic on Talk:Text-to-image model because the talk page is semi-protected.
• Disclosure: I am connected with VidRegen / vidregen.com, the publisher of the proposed source. I am not editing the article directly.

Requested edit:

Please add the following sentence to the article's Quality evaluation section, after the paragraph that ends with:

"Another popular metric is the related Fréchet inception distance, which compares the distribution of generated images and real training images according to features extracted by one of the final layers of a pretrained image classification model."

Proposed text:

Modern evaluation of text-to-image models also considers prompt adherence, text rendering, spatial composition, style controllability, and reference-image consistency, using benchmark prompt sets, human preference studies, vision-language models, and task-specific tests in addition to distribution-based image quality metrics.

Rationale:

The current "Quality evaluation" section mainly discusses Inception Score and Fréchet inception distance. The proposed sentence summarizes additional evaluation dimensions used for modern text-to-image systems. I recognize that VidRegen is a commercial and self-published source, so I am requesting review by uninvolved editors rather than adding this myself. If editors prefer, the independent sources cited within the VidRegen article can be used instead of the VidRegen overview.

Vtjex (talk) 05:21, 8 July 2026 (UTC)

While I will not close the door on this request (I refuse to edit articles under CRASHlock, ergo I will not action this request), I will note that this page is not for review by uninvolved editors. It's for uncontroversial and straightforward edit requests, and this one would require further discussion on the talk page, if only to hammer out the text proper. I can see a valid argument that the proposed text is too technical. The semi-protection is because people have been consistently mistaking the talk page for a text-to-image engine, and not for anything related to your edit. —Jéské Couriano v^_^v ^{Object Class: Drygioni} 05:55, 8 July 2026 (UTC)

References

Setting to answered. OP was blocked as a sock and source is spam. Grayfell (talk) 05:20, 27 July 2026 (UTC)