JJJ Facet E5 (old name: JJJ E5 Prompt Accurate) is a character-oriented LoRA with strong clothing, prop, and scene readability, substantial medium response, and a clear affinity for mature adult characters.
It works well for quiet character introductions, tired or lived-in portraits, structured outfits, workshop scenes, full anthropomorphic characters, and selected painterly or graphic treatments. It can keep a subject and environment readable while changing the rendering medium.
The listed use cases reflect the character and diagnostic test pack. Facet can also follow explicit palette, lighting, and medium directions: daylight, pink-blue neon, saturated pop, monochrome ink, and graphic treatments are all viable in the reviewed images. Warm amber-brown grading is common in some interior portraits, but it does not define every output or restrict the range of subjects you can try.
Precision depends on the requested feature and medium. The current review supports practical detail retention in many scenes, but it also shows age drift and specific instruction failures. It should be checked carefully when exact age, subtle hybrid traits, chest binding, facial completeness, or hand contact are essential.
Strengths
Readable layered clothing, accessories, props, and environmental cues.
Convincing mature presence, fatigue, and ordinary adult character expression.
Useful male, masculine, and androgynous character handling.
Strong full-anthro workwear and separation of fur, fabric, leather, and metal.
Responsive watercolor, ink wash, painterly, impasto, graphic, and fashion-illustration directions.
Prompt-responsive palette and lighting, including daylight, neon, saturated pop, and monochrome treatments.
Useful local detail following, including visible athlete abrasions and clear scene props.
Weaknesses
Explicit ages can be overshot, especially with white hair, fatigue, or workshop cues.
Some clothing and anatomical instructions are freely reinterpreted. Plain sleeveless tops can gain hoods; subtle tiger traits can become full anthro; bound-chest instructions can become exposed cleavage.
Strong gouache simplification can omit eyes.
Neon and close-up skin can remain glossy, and dark fabric can become overly lacquered.
Related prompts can return recurring facial types. Warm, dark interior rendering is common in some quieter portrait prompts; explicit palette, lighting, and medium cues can change the treatment.
Detailed fingertips do not guarantee correct hand origins, pressure deformation, or complex contact.
Risograph grain and palette may coexist with excessive realistic modeling.
Best use cases
Mature or tired adult character portraits.
Quiet desk, workshop, scholar, uniform, and formalwear scenes.
Full anthropomorphic mechanics and workwear characters.
Semi-realistic character illustration.
Watercolor and ink character studies, hard-block graphic portraits, and fashion illustration.
Pairing note
Note when pairing with my Chimera/Muse/Poise/Patina (with 0.8): at 0.25, JJJ Facet E5 starts behaving less like a competing style driver and more like a correction layer. It can improve prompt adherence and sometimes pull an overly aged or severe face back toward the requested presentation, without overpowering the primary LoRA as strongly as it did at 0.5. (So I recommend trying in the range below 0.50)
The pairing note records my mixing observations. The reference review below tests this LoRA alone at weight 1, so it does not establish the same behavior at 0.25 in a mixture.
Base models and reference settings
Basic Model: rRealisMAX Krea-2 - V1.5 or Krea-2 - Turbo. Other KREA2-derived bases can be tried experimentally.
Tested reference: rRealisMAX V1.5, LoRA weight 1, Euler, simple scheduler, 8 steps, CFG 1, VAE ae.sft, 768 x 1152. No trigger word was included in the reviewed prompts.
Earlier creator settings: er_sde, simple scheduler, 8-10 steps, CFG 1. Those settings were not tested in this review.
The review covers 66 prompts with two images each. Comparisons with my KT versions use different bases and seeds; the observations describe practical results rather than controlled LoRA superiority. Exact small-marking placement, precise iris-color consistency, and prescribed multi-layer background gradients still need dedicated testing.