Skip to main content
Participant
November 28, 2024
Question

Firefly ignoring prompts

  • November 28, 2024
  • 63 replies
  • 7225 views

Not really a bug, but I dont know what else to put it under. Firefly ignores most of my prompt and gives me simple low quality images that dont even make sense. I describe what I want with great detail and its frustrating that it ignores what I want and gives me random stuff. I heard great things about Firefly but so far it has beyond disappointing.

63 replies

Participant
August 15, 2026

Firefly is embarrassingly bad. The fact that a single person at Adobe thought this was ready for release is a testament to their universally poor quality control and utter laziness. It is consistently the least accurate, most frustrating AI I’ve used by an astronomically wide margin. It seemingly ignores prompts entirely and just does whatever it feels like. Most recently I told it to replace certain words with other words on a graphic, and it deleted all of the text and added pyro effects. The fact that it you have no recourse to refund credits, even when it completely whiffs, is a joke.

Tarun Saini
Community Manager
Community Manager
August 17, 2026

Hi ​@default37r5jfbyh9je,

 

Thank you for taking the time to share your feedback. I'm sorry to hear that your experience with Firefly has been so frustrating. We understand how disappointing it can be when the generated results don't align with your instructions, especially when credits are consumed in the process.

To help us better understand what happened, could you share a few additional details?

  • Which Firefly feature were you using (Generate Image, Generate Video, Text Effects, Generative Fill, Edit Image, etc.)?
  • Were you using the "Replace" action on an existing graphic or creating something from scratch?
  • Could you share the exact prompt (or a simplified version) that was used?
  • Do you have before-and-after screenshots showing the original graphic and the result that added the pyro effects?

 

Regards,

Tarun

Kartika Rawat
Community Manager
Community Manager
July 20, 2026

Firefly Image Model 5 (October 28, 2025) improved prompt adherence, including the missed-element issues raised in this thread. Many of the images here were generated on earlier models, so it's worth running the same prompt on Model 5 to see the difference. Thanks for documenting this in detail — that specificity is what helps the team target fixes.
If a prompt isn't landing, the Firefly AI Assistant can suggest a model and phrasing to match what you're going for: https://helpx.adobe.com/firefly/web/firefly-ai-assistant/firefly-ai-assistant-overview.html. You can also switch to one of the partner models available in Firefly (like Google's Gemini/Nano Banana Pro or FLUX.1 Kontext) if one of those is a better fit for what you're trying to generate: https://helpx.adobe.com/in/creative-cloud/apps/generative-ai/non-adobe-models-in-adobe-products.html

And if a result still misses, the thumbs-down button sends it straight to the team building the models.
 

peterk28060461
Inspiring
July 19, 2026

-

jeacret
Participant
July 17, 2026

I’m consistently amazed at how terrible it STILL is. It’s only helpful without prompts for removing elements. I tried to add a very simple red marker to a photo of a page. I just wanted to have it added with an appropriate shadow. I drew a shape the size of the marker at the angle I wanted. My prompt was something like “Add a thin-tipped red, correcting marker” Of the 3 attempts, 2 were just copies of the page with no change. The third was part of a red line? I tried again with something simpler “add a red marker”. This time, 2 more copies of the page below, but the third attempt had a red marker! However, it was just the first inch or so of a marker and then it was cut off. it had no shadow or anything else that would place it in the scene. This has been the type of experience I’ve had literally every time I forget how much firefly sucks and I try to use it. 

Not new here
Tarun Saini
Community Manager
Community Manager
July 20, 2026

Hi ​@jeacret,

 

We're sorry to hear about your experience and understand how frustrating it is when simple edits don't produce the expected results. We'd like to investigate further. Could you share the original image, the exact prompt(s) you used, and a screenshot or screen recording of the generated results?

 

Also, please let us know which browser and operating system you're using. This information will help us better understand whether the issue is related to prompt interpretation, object generation, or the editing workflow itself.

 

Regards,

Tarun

Participant
June 29, 2026

THIS!

All I wanted to do was take 2 pieces of jewelry and make a life style with a model I had to reup my credit so far

SOOOOOOOOOOOOOOOOOOO man attempts where it just outright ignored the reference
a few where the hands were backward!?!?
it started adding things from a previous prompt for a completely different picture with a completely different reference image
constantly changing the reference image
the few times it did follow the reference image it decided to add other jewelry out of no where

EDIT - now I’m getting disembodied hands!?!?!?


It’s like it’s trolling me and if this were say a outsourced freelancer they wouldn’t be getting paid (or credits) till they actually followed instructions let a lone finished the job, yet here we are

Tarun Saini
Community Manager
Community Manager
July 2, 2026

Hi ​@Secret_Asian_Man76,

 

Thank you for sharing your experience. We understand how frustrating it can be when you're trying to achieve a specific creative outcome and the generated results don't accurately follow your references or instructions.

To help us investigate further, could you share:

  • The prompt(s) you are using
  • The reference image(s) for the jewelry
  • A few examples of the incorrect outputs you're receiving
  • Whether this is happening consistently with all projects or only this specific one

We’re here to help, just need some info.

 

Regards,

Tarun

Ernest38070308idec
Participant
April 27, 2026

Поверніть гроши за цей непотріб!!! жодна генерація незважаючи на техничну якість не відповідала ментальної!!!

Tsukoden
Participant
April 18, 2026

+1 to this post. I excitedly tried it out having big hopes for an Adobe product and was significantly let down. After spending credits to create videos, the results were awful animations which ignored pretty much all of the prompt inputs with the exception of at least starting with the initial image I uploaded. Sadly, this is not ready for production and I will not be wasting anymore credits until I hear better results from their product. Hopefully one of the companies will come up with a reliable video generator since Sora is now end of support as well.

Participant
April 5, 2026

FIrefly is so embarrassingly bad, I cant believe ADOBE would release it. They deserve to go under just because of how bad it is. Shameful. WTF is wrong with them, and who decided to release this?

April 6, 2026

Hey ​@MindTickle 

Your opinion is valid. Depending on what you want to create Firefly may not meet your requirements.
Firefly is commercially safe as it uses it resources from Adobe Stock.

Firefly is great for ideation and to assist with getting a concept in motion w a client(s).
Adobe empowers artists to create something and then take it into one of the many apps to add the artist’s own creative flare, essentially take ownership of the creative process.


Whereas other models may not be commercially safe using content from other resources.

Generating images isn’t exactly one and done.
Cheers

Nate
 

Ondřej30637669tay4
Participant
March 19, 2026

I have really terrible experience trying to generate any video. It always does error, once it went through and generated totally unusable basic video which was typical when this all ai generation started. I use nim video and higgsfield and they are superrior agains this. My question is why? Why company with such budget can not offer serious AI tool? Even ARTLIST is better in that. Why is that? Are you working on improving? 

April 7, 2026

I find using simple prompts works best for Adobe Firefly. 

Chat GPT prompts are confusing and sometimes don’t make sense for Firefly.

ARTLIST might be “Better” keep in mind we don’t know where they get their reference material from; where as Adobe using Stock images from Adobe Stock: making it commercially safe.

Serious tools: Photoshop, Illustrator, and other are very powerful tools to create. The point of using Adobe tools is for us as artists to create our own creation.

Prompting is not exactly one and done.

cheers

Nate

Drew Myers 2026
Participant
March 14, 2026

A little honesty here. I realize this is Beta. I'm not sure why you even have this available when there are at least 6 or 7 apps much better than your beta. Sorry, I'm out.
 Takes forever to generate and does not not follow prompts.