Morphing is a crazy magician. She takes your prompt, finds a few words she likes, and cranks the volume up. Way up.
Sometimes you get something you didn't know you wanted, and you fucking love it.
Sometimes she surprises you so much that you're staring at the screen going, "Fuck you, man. This is wasting my time."
The body? The silhouette? The face? The whole damn vibe? She can get very interested in a few parts of your prompt and let them take over the show.
Turn her down when you just want a little help. Give her more room when you're in the mood to see what she's about to pull out of that hat.
What she tends to do
Can make particular body traits, shape contrasts and silhouettes read more strongly. In my old-prompt retests, specified fullness and robust proportions sometimes became much more prominent.
Can also push the face, pose, camera distance and rendering style along with them. A bigger-looking foreground arm may include perspective, not just a different body.
Has useful results well beyond body-focused characters: mechanical toys, material studies, elaborate clothing, food, paper craft, ordinary animals and environment illustrations all have examples in the reviewed batches.
Often brings a more polished or more opinionated character interpretation when the medium is left open. Some prompts stay fairly close to one idea; others give her more room to wander.
What she doesn't keep quiet about
She may amplify one trait while making another harder to keep. A fuller body can arrive with a younger face, stronger glamour or a closer camera. Mature age, restraint and masculine presentation don't necessarily get louder at the same time.
If you're reaching for her because a body type isn't reading clearly, she's worth a try. Just check what else changed. These are observations from standalone use, not proof that she'll act as a body-only helper in every LoRA mix.
How to use her
The main comparison galleries use strength 1.0. Think of that as a look at how much she might do, not a requirement for normal use.
Start lower when you want some help without handing over the whole image. Push her higher when you're specifically looking for a stronger interpretation or a surprise. I haven't established one perfect low-strength range, and her response depends on the prompt.
For a quieter starting point, try JJJ · Sleight. For a stronger drawn treatment with a different balance, try JJJ · Illusion.
Settings
Tested base: rRealisMAX Krea-2 V1.5-aio
Sampler / scheduler: Euler / beta
Steps: 8-10; the main comparison galleries use 10
CFG: 1 Strength Main reviewed batches: 1.0; lower strengths are worth trying for everyday use
Trigger: The reviewed prompts did not include jjjmosaic; you can start without it
Negative: Left empty in the reference runs
VAE: Automatic
Reference size: 768 x 1152
The 8-step starting point comes from my additional use. The structured comparison batches use 10 steps and strength 1.0; they don't establish an exact lower-strength sweet spot. More steps aren't the same thing as giving her fewer opinions.
About this version
This is the E10 checkpoint from the same mixed-dataset training as JJJ · Sleight (E4) and JJJ · Illusion (E7), released as another way to use that training rather than a replacement for both.
The structured review covers 87 prompts for this checkpoint. The additional old-prompt retests are practical examples, not a controlled proof of a body-morphing mechanism. The name is the vibe. Enjoy the useful surprises, and keep an eye on the details you actually need.