Using AI relationship prompts to shape Stable Diffusion concept brainstorming
I’ve been trying a method where I use structured prompt brainstorming to clarify ideas before generating images.
Focusing on narrative and emotional cues helps refine visual concepts like mood and character expression.
Breaking prompts into smaller denoscriptive parts seems to improve composition and detail in outputs.
It’s been interesting to see how organizing ideas textually influences the end result.
Curious how others prepare concepts before feeding them into generation pipelines.
https://redd.it/1rao5jk
@rStableDiffusion
I’ve been trying a method where I use structured prompt brainstorming to clarify ideas before generating images.
Focusing on narrative and emotional cues helps refine visual concepts like mood and character expression.
Breaking prompts into smaller denoscriptive parts seems to improve composition and detail in outputs.
It’s been interesting to see how organizing ideas textually influences the end result.
Curious how others prepare concepts before feeding them into generation pipelines.
https://redd.it/1rao5jk
@rStableDiffusion
Reddit
From the StableDiffusion community on Reddit
Explore this post and more from the StableDiffusion community
Just returned from mid-2025, what's the recommended image gen local model now?
Stopped doing image gen since mid-2025 and now came back to have fun with it again.
Last time i was here, the best recommended model that does not require beefy high end builds(ahem, flux.) are WAI-Illustrious, and NoobAI(the V-pred thingy?).
I scoured a bit in this subreddit and found some said Chroma and Anima, are these new recommended models?
And do they have capability to use old LoRAs? (like NoobAI able to load illustrious LoRAs) as i have some LoRAs with Pony, Illustrious, and NoobAI versions. Can it use some of it?
https://redd.it/1rambdn
@rStableDiffusion
Stopped doing image gen since mid-2025 and now came back to have fun with it again.
Last time i was here, the best recommended model that does not require beefy high end builds(ahem, flux.) are WAI-Illustrious, and NoobAI(the V-pred thingy?).
I scoured a bit in this subreddit and found some said Chroma and Anima, are these new recommended models?
And do they have capability to use old LoRAs? (like NoobAI able to load illustrious LoRAs) as i have some LoRAs with Pony, Illustrious, and NoobAI versions. Can it use some of it?
https://redd.it/1rambdn
@rStableDiffusion
Reddit
From the StableDiffusion community on Reddit
Explore this post and more from the StableDiffusion community
Free SFW Prompt Pack — 319 styles, 30 categories, works on Pony/Illustrious/NoobAI
https://redd.it/1raq9a0
@rStableDiffusion
https://redd.it/1raq9a0
@rStableDiffusion
Reddit
From the StableDiffusion community on Reddit: Free SFW Prompt Pack — 319 styles, 30 categories, works on Pony/Illustrious/NoobAI
Explore this post and more from the StableDiffusion community
lora-gym update: local GPU training for WAN LoRAs
Update on lora-gym (github.com/alvdansen/lora-gym) — added local training support.
Running on my A6000 right now. Same config structure, same hyperparameters, same dual-expert WAN 2.2 handling. No cloud setup required.
Currently validated on 48GB VRAM.
https://redd.it/1ravptl
@rStableDiffusion
Update on lora-gym (github.com/alvdansen/lora-gym) — added local training support.
Running on my A6000 right now. Same config structure, same hyperparameters, same dual-expert WAN 2.2 handling. No cloud setup required.
Currently validated on 48GB VRAM.
https://redd.it/1ravptl
@rStableDiffusion
GitHub
GitHub - alvdansen/lora-gym: Open-source training pipeline for character identity, motion, aesthetic, and style LoRAs. Built on…
Open-source training pipeline for character identity, motion, aesthetic, and style LoRAs. Built on musubi-tuner. - alvdansen/lora-gym