Software Engineering Courses (SECourses)•6mo ago

can you train models using it?

grandtheftdeltaOP•6/26/25, 11:53 PM

how does it differ from flux.1

Ggrandtheftdelta can you train models using it?

Furkan Gözükara SECourses•6/26/25, 11:55 PM

well we need to wait kohya

grandtheftdeltaOP•6/26/25, 11:55 PM

what would the results be do you think? Better than Flux.1 or worse?

grandtheftdeltaOP•6/26/25, 11:55 PM

it seems to be strong model

Ggrandtheftdelta what would the results be do you think? Better than Flux.1 or worse?

Furkan Gözükara SECourses•6/26/25, 11:57 PM

not sure atm but this is a more stronger model

Furkan Gözükara SECourses•6/26/25, 11:59 PM

Furkan Gözükara SECourses•6/26/25, 11:59 PM

just prompt no masking or inpainting

Furkan Gözükara SECourses•6/27/25, 12:07 AM

FFurkan Gözükara SECourses Click to see attachment

AnonDreamer•6/27/25, 12:31 AM

are these in the model downloader im not seeing it

Pelolisu•6/27/25, 12:38 AM

Does kontext do LQ restoration like SUPIR?

Pelolisu•6/27/25, 12:42 AM

try do restoration from that band image

AAnonDreamer are these in the model downloader im not seeing it

Furkan Gözükara SECourses•6/27/25, 12:56 AM

yes it is

Furkan Gözükara SECourses•6/27/25, 12:57 AM

i made tutorial editing

PPelolisu Does kontext do LQ restoration like SUPIR?

Furkan Gözükara SECourses•6/27/25, 12:57 AM

it did a degree

Furkan Gözükara SECourses•6/27/25, 12:57 AM

Furkan Gözükara SECourses•6/27/25, 12:59 AM

0056001-make20the20man20has20big20long20beautiful20blac-flux1-kontext-dev-1.png

Furkan Gözükara SECourses•6/27/25, 1:00 AM

0056002-make20the20man20has20big20long20beautiful20blac-flux1-kontext-dev-3.png

FFurkan Gözükara SECourses Click to see attachment

Pelolisu•6/27/25, 1:03 AM

Maria Betania

Furkan Gözükara SECourses•6/27/25, 1:38 AM

i think i will publish tomorrow really late now

Furkan Gözükara SECourses•6/27/25, 1:38 AM

but exporting video sorry for delay

Darth Teethius•6/27/25, 4:28 AM

is there a hidream trainer out yet?

DDarth Teethius is there a hidream trainer out yet?

�

𝗡𝗶𝘁𝗶𝗻 𝗣𝗮𝗿𝗲𝘃𝗮•6/27/25, 4:32 AM

AI Toolkit, Diffusion Pipe, One Trainer and Simple Tuner

�

𝗡𝗶𝘁𝗶𝗻 𝗣𝗮𝗿𝗲𝘃𝗮•6/27/25, 4:32 AM

Only Kohya doesn't have it

Darth Teethius•6/27/25, 4:33 AM

thank you

Castor Troy•6/27/25, 4:57 AM

Do you guys think flux loras will ever work for kontext?

daroth•6/27/25, 6:09 AM

not in the official workflow

daroth•6/27/25, 6:27 AM

getting absolutely minimal effects when inserting lora loaders to the workflow

daroth•6/27/25, 6:57 AM

I expect we'll have lora use and training in only a couple of days

SnicksterJay•6/27/25, 7:58 AM

Sorry for the simple question but when I want to download an image I just generated in swarmui how can I choose a directory for the download to output to?

SSnicksterJay Sorry for the simple question but when I want to download an image I just genera...

J2_Nedo•6/27/25, 8:46 AM

right click on the generated image / open in new folder. But if you meant to save generated Images to a specific folder, you can change this in the settings in Output directory.

MetaKan•6/27/25, 11:54 AM

Hi, I saw that it's possible to train LoRAs for Flux Kontext dev. @Dr. Furkan Gözükara have you had a chance to check it out?

Mathew•6/27/25, 12:12 PM

@Dr. Furkan Gözükara
Hey Furkan, now with the new Flux Dev license (non-commercial, not even for outputs), are you thinking of switching to Flux Schnell or Chroma? Seems like the best way forward… we really need to boycott these kinds of moves from BFL

Inoshi•6/27/25, 1:16 PM

new attention https://github.com/mit-han-lab/radial-attention

GitHub

GitHub - mit-han-lab/radial-attention: Radial Attention Official Im...

Radial Attention Official Implementation. Contribute to mit-han-lab/radial-attention development by creating an account on GitHub.

MMetaKan Hi, I saw that it's possible to train LoRAs for Flux Kontext dev. @Dr. Furkan Gö...

Furkan Gözükara SECourses•6/27/25, 1:28 PM

nope

MMathew @Dr. Furkan Gözükara Hey Furkan, now with the new Flux Dev license (non-commerc...

Furkan Gözükara SECourses•6/27/25, 1:28 PM

nope

Furkan Gözükara SECourses•6/27/25, 1:28 PM

they dont care individuals how they use

Furkan Gözükara SECourses•6/27/25, 1:28 PM

licence is for SaaS

Furkan Gözükara SECourses•6/27/25, 1:29 PM

if you are FAL or Replicate then you need to pay money

IInoshi new attention https://github.com/mit-han-lab/radial-attention

Furkan Gözükara SECourses•6/27/25, 1:31 PM

sad not in official comfyui

FFurkan Gözükara SECourses they dont care individuals how they use

Mathew•6/27/25, 1:35 PM

Yeah, for sure they care more about the big players. But now, theoretically, if I sell an image I generated, I’d be doing it illegally, right?

Mathew•6/27/25, 1:38 PM

I’m thinking of testing out Chroma and Freepek having an alternative seems like a good idea

MMathew Yeah, for sure they care more about the big players. But now, theoretically, if ...

Furkan Gözükara SECourses•6/27/25, 1:45 PM

i doubt that

Furkan Gözükara SECourses•6/27/25, 1:45 PM

they know that individuals will use it. but they are making huge money from SaaS

Furkan Gözükara SECourses•6/27/25, 1:45 PM

this increases their reach

Inoshi•6/27/25, 2:14 PM

any prompt to blend a photorealistic character with a 3d rendered one in the same photorealistic scene for the Kontext?

潮

潮汐•6/27/25, 2:23 PM

Hello Furkan, I'd like to inquire whether the batch processing function does not support prompt words, or can only multiple pictures share the same prompt word?

潮

潮汐•6/27/25, 2:23 PM

Is there anyone kind enough to answer my question? Thank you very much.

FFurkan Gözükara SECourses they know that individuals will use it. but they are making huge money from SaaS

Mathew•6/27/25, 2:47 PM

Yeah, that makes sense, but it would be so much better if they made it fully open at least for individuals. There are so many people building tools and cool stuff with the dev version, and just the possibility of getting sued can be really discouraging for a lot of folks.

潮

潮汐•6/27/25, 3:22 PM

@Dr. Furkan GözükaraHello Furkan, I'd like to inquire whether the batch processing function does not support prompt words, or can only multiple pictures share the same prompt word?

Darth Teethius•6/27/25, 4:21 PM

Since chroma is schnell based, would it train with just the schnell training scripts or does anyone know if it needs a new one?