P: Text and symbol generation in images

[#1]

I am trying to generate a logo with a desired text. It is not accepting it.

[Roxanne32010064fw7y]

Thanks

[Rosane35216144s9ty]

Ele não gera textos, falei pra ele gerar um texto e ele gerou um cara com cabeça de tigre, um de aguia e varias paisagens nada aver

[droopydog500]

Please share the prompt you used and the resulting image for us to help you better. Thank you.

[Corinna35129583ycdw]

Hallo,
Ich schätze hier wurde Stil mit Stift verwechselt! Hahaha

[viatonx]

Hello. We get unsuccessful results when we want anything to be written on objects.
We are looking forward to more realistic and accurate results.

[PMJorg]

Every time, with no exceptions, when I ask Adobe Firefly to create 3D lettering on a gallery wall that spells out a phrase, I get this:

Here is my prompt: 

white 3D lettering on a white wall, lighted with spot lights from above and the left side that says (spell each word exactly like I wrote it in between " and "): "Reality as we see it" and "is a hallucination created by our brain" in two lines

Why is this? AI is highly trained on text, and knows how to spell.

[droopydog500]

While it has gotten better, Firefly (and most other AI image generators) is not good at putting exact text from a prompt into an image. Sometimes it is perfect, but usually, it does not get it right.  The best solution is to use Adobe Express or Photoshop to add text in.  The more complicated the angle or perspective of the text, the more likely you will need Photoshop to get what you want.

[PMJorg]

Maybe Firefly suffers from the same oversight or blind-sight as new artists and designers. When focusing most of its energy into  how something looks, rather than splitting the attention to detail between the visual and the content, it makes errors due to inattentiveness. Learning a  healthy balance between appearance and content takes effort at first, but eventually good designers learn it and rarely make this kind of mistake as they go forward. Firefly needs to use both eyes, with one on appearance, the other on content. 

[rayek.elfin]

GenAI is not intelligent: the algorithm doesn't actually understand what it is doing. The term "Artificial Intelligence" is a misnomer, and popularized, but there is no awareness there. It's not even 'stupid'. It is merely an algorithm and can't even create true random numbers (hence: the seed).

Comparing Firefly with a real human intelligent designer is like comparing a rock with a planet that has a vibrant natural ecosystem.

[DawnAble]

AI image diffusion models can be trained on text. Unfortunately, Adobe's Firefly models are not. I hope they rectify this soon. 

In the meantime, DALL-E 3 and Midjourney v6 will.  

[PMJorg]

Thank you for that information. I've noticed ChatGPT is no better at it. I'll have to try the other two!

Get Outlook for Android<>

[@mj]

Hi,

I won't sugarcoat it. Firefly is terrible at text.

I am aware this is on the product teams radar.

When the fix will come is not known.

Best

mj

iMSD

[briced40890199]

Je demande de générer une image où il y a écrit "Bon anniversaire" (en Français) et il me génère des images où il y a écrit "Happy birthday" (en anglais), ce qui est totalement stupide car je suis Français. Il va falloir apprendre à programmer...

[Anonymous37037930cx5u]

It is not able to generate correct text in images using the prompt.

[droopydog500 · Accepted solution]

@Anonymous37037930cx5u, While it has gotten better, Firefly (and most other AI image generators) is not good at putting exact text from a prompt into an image. Sometimes it is perfect, but usually, it does not get it right.  The best solution is to use Adobe Express or Photoshop to add text in.  The more complicated the angle or perspective of the text, the more likely you will need Photoshop to get what you want.

Please see this article on why generating text in AI image generators is hard:

https://theconversation.com/if-ai-image-generators-are-so-smart-why-do-they-struggle-to-write-and-co...

[Ahmet38651091lb88]

I needed an image with the word PARK on it to use in my game, but the application could not do this. He put trees around and the text was very small, and when he said give without trees, the same thing happened again. In another parking article, when I added it to my game, the image remained horizontal and did not work for me.

[Kartika Rawat]

Hi Ahmet38651091lb88!

Sorry about the trouble. Can you share what prompt you're using? Also, please post a screenshot that shows the issue. 

Let us know.

Thanks,

KR

moved to discussions.

[Kevin Stohlmeyer]

@Ahmet38651091lb88 No AI can produce legible text at this time - it's a known limitation for generative AI including firefly. Firefly also does not recognize negative prompts so "without trees" would produce trees. Describe the image you want without the work "park".

[Ahmet38651091lb88]

I think I solved my problem, I made the article that says PARK, can I use them in my unity game ?

[Manuj30113954tvzk]

Hi,

I tried creating a new logo for my company for trials and I realised that I even specifically gave name "ZEHEN" that too in capital to Adobe Firefly but it was not able to create that or add that. Also I tried creating a broucher but there was no option to add text. Infact I gave generative Ai mutiple commands to add specific text which I wanted but it was only generating art forms of same with incorrect spelling So kindly look into the same & improve the thing before releasing the paid version as this will be a major issue for most.

[alfred31246100tlof]

Cannot readily input correct spellings of words in to the image generator. 

[Nick313292751q2i]

Tried this out with various prompts - it reflected the language in my prompt each time, but the spelling was completely broken. 

[MANOJ322846159oko]

Getting spelling mistake while creating with text

[Preran]

I hear you. The tool is getting better at creating text but there is some way to go. Feel free to share your feedback using the options in Firefly. 

Thank you for your feedback.

[abhi32288682zpe8]

For example, for the prompt "A  lady wearing formal suit and listening to an audiobook summary named "the power of subconscious mind" , The image generated is contining the text for the book name as "Thie povf ppower sunisosndvee mind"