<h1 id="rockyphm-krea-2-lora">rockyPHM — Krea 2 LoRA</h1>A character LoRA for Rocky from Andy Weir's Project Hail Mary.The character belongs to Andy Weir. The visual design in this LoRA is my own interpretation — it is not an official or canonical depiction.Status: work in progress. It works well enough to be useful, but it has real limitations (listed below) and I'm still expanding the dataset. Feedback and failure cases are very welcome.<hr /><h2 id="trigger-word">Trigger word</h2><pre><code>rockyPHM
</code></pre>Use it to name the subject. The base model has no prior for this token, so the LoRA does all the work — without it you get nothing related.<hr /><h2 id="recommended-settings">Recommended settings</h2>Setting Value Base model Krea 2 (trained on Krea-2-Raw, works on Turbo) LoRA weight 1.0 Sampler <code>er_sde</code> + <code>simple</code>, 8 steps Alternative <code>euler</code> + <code>simple</code> Generation resolution 1024 tested and stable<hr /><h2 id="prompting">Prompting</h2>Name the subject with the trigger and let the LoRA handle the anatomy. Don't describe the carapace, the texture or the limb structure — that competes with the trained weights and usually makes things worse.Limbs — call them <code>limbs</code>. The creature has five, arranged radially: four at the sides and one at the rear center. If you get too many or too few, force it explicitly:<pre><code>standing on five limbs
</code></pre>Digits — call them <code>digits</code>. Three per limb tip. Same trick applies if the count drifts:<pre><code>digits spread open
three digits gripping
</code></pre>Scale — the LoRA has a built-in preference for the subject being fairly large in the frame. When there's a second subject, say so explicitly (<code>small</code>, <code>occupying a narrow portion of the frame</code>) or the two will come out the same height.<hr /><h2 id="what-works-well">What works well</h2><ul><li>Material substitution — glass, metal, anything. The structure survives; the LoRA learned geometry, not just texture.</li><li>Non-photographic styles — flat vector, hand-drawn illustration, poster art, children's book. Holds up in all of them.</li><li>Extreme lighting — backlight, silhouette, single coloured source from below. The shape stays readable and the light is built correctly rather than applied as a filter.</li><li>Contact with objects — gripping, holding, touching. Digits interact with objects plausibly.</li><li>Any environment — no background baked into the model. Indoor, outdoor, abstract, all fine.</li><li>Extreme perspective — fisheye and heavy distortion don't break it.</li></ul><hr /><h2 id="known-limitations">Known limitations</h2><ul><li>Limb count degrades when nothing touches the ground. Airborne poses, falling, tumbling — expect six or seven limbs. Keep at least a couple of limbs in contact with something.</li><li>The fifth limb is rear-center. In head-on front views it's occluded and often simply not generated. Use a three-quarter angle, or state it explicitly.</li><li>Two instances in one scene degrade the limb count on both.</li><li>Feature-space contamination. Placed next to boulders, rocks or mineral surfaces of a similar colour, the subject drifts toward them and loses its own texture. Specify a contrasting material for the surroundings, or separate them with depth of field.</li><li>Flat / vector rendering makes the limb count less stable than photoreal, because flat colour removes the shading that separates overlapping limbs.</li><li>Partial occlusion instructions are often ignored (<code>half hidden behind a crate</code> tends to put the subject on top of the crate instead).</li><li>Digit count occasionally drifts between two and three. Naming <code>three digits</code> helps.</li></ul><hr /><h2 id="example-prompts">Example prompts</h2>Photoreal, neutral:<pre><code>rockyPHM standing in a compact pose, seen head-on at eye level, full body shot, soft studio lighting, seamless gray backdrop
</code></pre>In a scene:<pre><code>rockyPHM standing on a rough wooden workbench, gripping a small metal tool with one front limb, three-quarter view, natural window light, blurred workshop background
</code></pre>Material substitution:<pre><code>rockyPHM made entirely of clear polished glass, transparent body with light refracting through the limbs, standing on a white surface, studio lighting
</code></pre>Illustrated style:<pre><code>rockyPHM standing beside a massive stone statue, reaching up with one front limb, flat 2D preschool television animation style, thick clean dark outlines, simple rounded shapes, bright saturated colors, minimal cel shading, side view
</code></pre><hr /><h2 id="training-details">Training details</h2>Base model: Krea-2-Raw Network LoRA: rank 32 / alpha 32 Training resolution: 768 Steps: 2800 Dataset: 26 imagesThe dataset was generated image by image with ChatGPT — slow, manual work, and the main reason the model still has the rough edges listed above. Keeping a single consistent design across separately generated batches turned out to be the hardest part of the whole project.<hr /><h2 id="feedback-wanted">Feedback wanted</h2>The dataset is the bottleneck. Specifically useful to me:<ul><li>Failure cases — prompts where the anatomy breaks, especially ones not already in the limitations list.</li><li>Generated images you're happy with that I could use to expand the dataset (let me know if you'd rather I didn't).</li><li>Sampler and weight combinations that work better than the ones above — I've only tested weight 1.0.</li></ul>

Files with SHA256: 63f505a5619b6904ddf576e989bb02d539145ed6f0fd2edb323506f15baabd02