firefly-ai-assistant

[inner ore]

did it last night prompt jockey event

did a experiment like Risograph style poster using custom model generated on Adobe Firefly

[frail narwhal]

Some things that I’d like to see:

-Speech recognition integrated (ex. ChatGPT’s speech recognition is good)
-The ability to lower the sycophancy. (I like compliments as much as the next guy but am uninterested to see them from this)
-Being able to access the prompts/tools it uses for future reference and tweaking. (ex. How you can “Load” in boards is great)

But this is a cool and an exciting new feature, I’m only nitpicking assuming it’s desired. I’ve noticed some of the “A few things worth noting for a potential next iteration” and “I can do this next” were ideas that were right on the money and in the exact direction I was wanting it to go.

[wise iron]

Hey all, I got the invite for the beta, and when I clicked in I still get the notice to sign up for the beta. Anyone else have this issue?

[polar pier]

@wise iron It is being rolled out. I have access in Germany.

[wise iron]

Ah, OK. The email made it sound like I was good to go. I'll give it some time. Thanks.

[polar pier]

@wise iron Are you on an Interprise plan?

[wise iron]

Enterprise with my company, yes.

[polar pier]

@wise iron You'll need to contact your admin to turn it on.

And allow access to it.

[wise iron]

Well that's good to know - I'll contact IT. Thanks again.

[polar pier]

Good luck!

[viral ermine]

@wise iron Shawn is correct, your admin should be able to enable access to the assistant. However, I wanted to note that there are a handful of enterprise plans that cannot access the assistant. Here's a list of who can and cannot access the assistant (link here) in case it is helpful to have when you reach out to your admin.

[frail narwhal]

Some things I think would improve this:

1. The ability to remove images etc. from the chat would be beneficial. (To clear clutter and reduce the potential of steering future generations in an undesirable direction.)

2. The option to open multiple images at a time with PS Web/Boards as well as a PS Desktop option. (To select multiple generations in the same manner you can reference multiple ones by clicking on the and the checkbox checks)

3. Sometimes it will just generate things on it’s own. Example: I generated a necklace, then I generated a mannequin for necklace mockups and as soon as the mannequin materialized it started a third generation with the mannequin wearing a necklace automatically. That was the direction I was wanting to go but what if I was unsatisfied with the mannequin and also not being able to control the credit expense. (This automatic thing happened in other instances as well). Every time it was doing something appropriate and in line with where I was going but it was ultimately unwelcome and not asked for.

4. (I later saw you can easily click and drag the images I refer to in the following) To be able to click on reference images that you previously referenced that are shown in the chat as a message from the user, on the right. To make it clear, what I circled in the picture below, was a generated image I used as a reference in chat in a prior message and is not clickable/selectable and to reference it again in the chatbox I’d have to scroll higher to the one the assistant originally generated. In many situations I could imagine the amount of scrolling being a headache. In fact maybe if there was little tiny thumbnails somewhere on the edge that enlarge when you hover over them to easily use them as an image reference and/or jump to that part of that chat would be convenient. There where multiple times I attempted to reference an image I recently “sent” or “referenced” as it seems natural but instead scrolled higher to the original.

[viral ermine]

Thank you for this thorough rundown! It’s super helpful. I’ll make sure the team gets this feedback!

[frail narwhal]

I can't think of a better response it could have given.
I asked both ChatGPT and Gemini to do the same thing and they both generated a black and white image. Claude made me a black and white SVG. All four of them complimented me in some way.

[whole rampart]

I have notice there is something called Compositing in the Productions section. I try to click on the new Beta there and it logs me out.

[wise iron]

Thank you - I'll pass this along.

[viral ermine]

What do you think about the compliments? Curious to hear as we continue to work on the assistant. Is it overdoing it? Would you prefer a more neutral response? Or do you like them?

[frail narwhal]

I would 100% turn them off the compliments and get to the point faster if I could. It's almost like I'm putting numbers into a calculator, press the equal sign and "nice numerical choice" pops up on the screen before it gives me the answer.
But I think if my older neighbor or my Aunt used this they would enjoy the compliments.

[whole rampart]

I would choose no compliments

[viral ermine]

@frail narwhal @whole rampart That's super helpful to hear. We've heard similar sentiments from others. A grating personality would be a big roadblock to using the assistant, so getting the personality right is crucial. Appreciate you weighing in on this.

[viral ermine]

Sorry I missed this yesterday! Would you happen to be able to share a screenshot of exaclty what you're doing you're doing right before you get logged out? Something clearly up, because when I try to use the assistant in FF production, it just "thinks" and never does anything. But I'm not getting the log out behavior so I'd love to have screenshot for when I report.

[hushed charm]

fwiw i am able to click on it production>composite but it loads forever and then eventually fires me back to the main firefly homepage

[viral ermine]

Gotcha. I will report this!

[frail narwhal]

For anyone interested in using speech recognition windows comes with “voice typing” you can conveniently access pressing the windows key and h.
You might have to do so some first time setup as far as mic permissions etc. I didn’t have to but I have a feeling I did so already in the past.

According to ChatGPT: Mac has one called “dictation” that works by pressing the Fn key twice or sometimes the mic key

[ivory abyss]

hi everyone, is it too crazy to get a " Just in case / save here/ ( WE ARE IN A GOOD SPOT)" instead of creating a new conversation, and maybe a transcript button or speech to text in the situation you need to describe or give more detail?

Something happened to my project, I think I broke it, and I can't get the assistant to respond or get some how the media to load into a new assistant (10 hours back and forth)... implementing in other ai refs cleaning some in photoshop to lock a look.... do they have a chat limit? am i doing it wrong? (mac m1) using only one assistant to develop a short idea, for ai vfx class.
I checked the other assistants, they do work, weren't working last night, only the one i put my weekend in, is the one that gives me this text :/

thank you and sorry for texting on the weekend,
hope yours is going great 🙂

[hushed charm]

you want to save a specific chat and come back to it? what was the last thing you did before it broke?

[ivory abyss]

more like saving a specific output from the agent where you can split to another idea or if you went down a rabbit hole you can go back to this point without confusing the assistant.

when that assistant stoped working, I was gettting to develop characters based on references given to the assistant, since i was waiting for credits to renew, so i was trying and doing test or everything else on my side to in order to advance a bit while I waited, then the time to use credits was ready and the agent it kept in a loop where I ask anything, tries to load and keeps bringing the "somethign went wrong please try again"

thank you 🙂

[hushed charm]

I don't think the credits resetting would affect it, there was a creative cloud outage today and that could have been causing issues.

it looks like it's still ongoing so you might need to just wait it out https://status.adobe.com/

[viral ermine]

Nothing is too crazy to ask for. We really want to hear what you want and what it would solve for you. I'll share this feature request with the team. I'll also run the issue you ran into by them as that sounds like something we need to fix beyond just offering a feature that would help with future hiccups. Also, feel free to post in here at anytime. We'll be quickest to reply during weekday business hours, but you're welcome to drop your thoughts in here at anytime.

[frail narwhal]

I would like to select/drag things into the chatbox from one chat, and go into another chat with that same thing still in the chatbox available to use as a reference. Example: had a chat making a logo in the past and want to use that logo and have it appear in a collage in my most recent chat.

Additionally, I noticed that only 1 reference image is allowed for video generation. I know it uses Veo 3.1 Fast and that using a first and last frame reference is possible in firefly. I would like to be able to do that with the assistant.

Also if speech recognition is being developed, I think it’s important to be able to continue talking while navigating away from the chat box window. Example: start talking and then simultaneously browse your computers files for a reference image.

[verbal solstice]

Hi AI, can I upload 4-5 images to be used as guides for the genai of the image? It looks like firefly only allows 1 image uploaded, is that right?

[hushed charm]

#community-intros message

[verbal spindle]

Fahhhhhh

[void ledge]

Hey @verbal solstice - firefly models can accept a style and composition reference image. some of our partner models accept up to 6 reference images. you might also look into firefly "custom models" where you can upload a set of 10-30 images and train your own, personal model to generate consistent images in that style.