Danrisi

UltraReal Fine-Tune Anima v1 release U can check it huggingface.co/Danrisi/UltraR… here and here civitai.red/models/2585622… Also some example on reddit reddit.com/r/StableDiffus…

@ClementsCopy Good question. I trained edit models for qwen-edit to transform anything into realistic, but didn't release it. Now I want to retrain it specially for flux Klein edit

@DanrisiTech Have you messed with Editing inside Flux Klein, Danrisi? Im finding it very tough to train for

My main loras now for Flux.Klein9b and ZImageBase All of them you can find on my HF huggingface.co/Danrisi and Civit civitai.com/user/Danrisi also more examples on reddit reddit.com/r/StableDiffus…

Flux2 LoRA is coming

Here is the release of my new LoRA that takes you back to the 2000s point-and-shoot era. ​I trained it entirely on my own photos using a Nikon Coolpix S4 and Canon PowerShot A570 to capture that authentic vibe. Hope you enjoy it ​civitai.com/models/2163880

Cooking a new lora (Actually 2, another one for qwen-edit)

2000s

God sometimes i forgot how much of a sensory overload this game is

@IamEmily2050 qwen with lora

Seedream 4 This is peak Gotham energy { 2004 VGA bar-selfie: Joker (smudged white greasepaint, green-tinted slicked hair, purple satin shirt open to chest, lit cigar) holds flip-phone at arm’s length, wide-angle lens slightly tilted. Batman (black cowl, matte finish, visible jaw stubble, grey T-shirt) sits centre, eyes narrowed at lens, one brow raised. Catwoman (black PVC halter, cat-ear headband, smudged eyeliner, red lipstick) leans over bar, gloved hand on Joker’s shoulder. Harley Quinn (red/blue crop top, diamond face paint cracked, pigtails with faded ribbon) pops between them, tongue out, holding a half-empty beer bottle. Background: dim wood-paneled dive bar, Bud Light neon blur, CRT TV static, jukebox glow. Harsh on-camera flash blows highlights, green-yellow white-balance shift, heavy VGA noise, 640×480 pixel stretch, date-stamp ‘04-10-15 02:17’. Mild motion blur on Harley’s bottle, dust specks on lens, finger partially covers corner. --ar 4:5 --style raw", "style": "photographic 2004 VGA analog selfie", "negative_prompt": "logos, text, extra limbs, smooth skin, HDR, modern phone", "output": { "format": "jpg", "long_edge_px": 1536 } }

@REVOLVO_OCELOTS @toyxyz3 did u even tried loras for qwen edit? or use non-euler_simple generation settings?

@toyxyz3 I wish qwen edit was good as this

This time I decided to train a LoRA for Qwen-Image to capture the nostalgic 90s / Y2K movie aesthetic. You can go make your own Blockbuster-era film stills. You can find lora here: huggingface.co/Danrisi/Qwen_9… and here: civitai.com/models/1950672…

@AIWarper Author used another model huggingface.co/Danrisi/Lenovo… I can recognize some motion blur and overexposed light elements I trained

Link to the actual Lora and the creator of it x.com/multimodalart/…

@DkdCool58445 @venturetwins @grok @gork If your toaster has at least 3080 then it will be fine. Anyway, I test on glif and it's working okay. I just like to generate locally or on cloud compute to set a lot of steps and some weird samplers to get maximum quality 😅

Blown away by the Qwen ultra realism LoRA 🤯 It makes candid photos that look like they were taken on an old iPhone, from really simple prompts. Had to use it for a "Capybara Goes to College" series.

reddit.com/r/StableDiffus… Made on reddit a small post with some images I got with Qwen-image and my lora

@umang_ai I'll DM you

@DanrisiTech I checkout your handwritten text LoRA and really impressed by your work. I would love to have a quick word with you around the same, can we do that?

Cooking up a cleaner and less blurred LoRA — reworking my Samsung dataset for Flux so it vibes better with Qwen-Image P.S.: image made not with final version of LoRA

@vamshi_nenu @angrypenguinPNG AIILF

@angrypenguinPNG EILF, E Girl I’d like to Is this the end? Of GILF’s ? 😂😂🙆🏽‍♂️

Instagirl LoRA v2.3 for Wan 2.2 is here. Open-source AI video is getting ridiculous 🤯

@DoggyDog1208 @_JKNFT_ @angrypenguinPNG I don't have ferrari but looks like f430

@burninganna @FAL @krea_ai @jfischoff You should try fine-tuning Qwen — it gives the best character consistency

Fine-tuned the FLUX.1 model with @FAL and @krea_ai at the Fal offices (I was just listening haha). Thanks @jfischoff for hosting - great night with the crew!