logo
welcome
Ars Technica

Ars Technica

Hobbyists discover how to insert custom fonts into AI-generated images

Ars Technica
Summary
Nutrition label

81% Informative

AI image generator Flux is capable of rendering depictions of accurate text.

Flux can now insert words rendered in custom fonts into AI image generations.

LoRA (low-rank adaptation) is a technique that allows users to augment knowledge in an AI base model with modular add-ons that have been custom-trained.

As of this writing, there are only two custom Flux typeface LoRAs, but we've already heard plans of people creating more as we write this. While it's still in its earliest stages, the technique of creating typeface LoRAs may become foundational if AI image synthesis becomes more widely deployed in the future. Adobe, with its own image synthesis models, is likely watching..

VR Score

79

Informative language

77

Neutral language

42

Article tone

informal

Language

English

Language complexity

52

Offensive language

not offensive

Hate speech

not hateful

Attention-grabbing headline

not detected

Known propaganda techniques

not detected

Time-value

long-living

External references

24

https://www.reddit.com/r/StableDiffusion/comments/1eyu7vj/comment/ljfwly3/https://github.com/aredden/flux-fp8-apihttps://cdn.arstechnica.net/wp-content/uploads/2024/08/hamburger_car-640x366.jpghttps://civitai.com/models/196131/spider-man-xl-sd15https://civitai.com/models/667307/flux-y2k-typefacehttps://cdn.arstechnica.net/wp-content/uploads/2024/08/now-that-we-have-flux-it-is-possible-to-train-text-typeface-v0-g1nx2nnioujd1-640x423.jpghttps://www.reddit.com/r/StableDiffusion/comments/1ex0yg1/comment/lj2mwy0/https://www.reddit.com/r/StableDiffusion/https://civitai.com/models/662069/chalk-board-drawing-fluxhttps://www.reddit.com/r/StableDiffusion/comments/1ex0yg1/now_that_we_have_flux_it_is_possible_to_train/https://www.reddit.com/r/StableDiffusion/comments/1ex0yg1/comment/lj2ire6/https://cdn.arstechnica.net/wp-content/uploads/2024/08/now-that-we-have-flux-it-is-possible-to-train-text-typeface-v0-bapi5jzgoujd1-640x284.jpghttps://arxiv.org/abs/2106.09685https://cdn.arstechnica.net/wp-content/uploads/2024/08/now-that-we-have-flux-it-is-possible-to-train-text-typeface-v0-ol4tpwuhoujd1-640x466.jpghttps://cdn.arstechnica.net/wp-content/uploads/2024/08/without_with_2.jpghttps://cdn.arstechnica.net/wp-content/uploads/2024/08/hamburger_car.jpghttps://cdn.arstechnica.net/wp-content/uploads/2024/08/now-that-we-have-flux-it-is-possible-to-train-text-typeface-v0-ol4tpwuhoujd1-scaled.jpghttps://www.reddit.com/r/StableDiffusion/comments/1ex0yg1/comment/lj4nu6j/https://cdn.arstechnica.net/wp-content/uploads/2024/08/custom_fonts_car-640x366.jpghttps://cdn.arstechnica.net/wp-content/uploads/2024/08/queen_of_universe-640x366.jpghttps://cdn.arstechnica.net/wp-content/uploads/2024/08/theking-640x366.jpghttps://www.reddit.com/r/midjourney/comments/10iixwp/anyway_to_create_an_image_of_a_person_wearing/https://civitai.com/models/674148/cyberpunk-typefacehttps://www.reddit.com/r/StableDiffusion/comments/1eyu7vj/comment/ljijozi/

Affiliate links

no affiliate links