Rebaked a WAI-Anima character LoRA onto upstream Anima-Base with off-distribution Gemini data. Trigger-only usable, face fidelity beats v1, intakes still cap out.
Tested on M1 Max 64GB ComfyUI: one character LoRA across 6 Anima derivatives, same prompt/seed. Trigger-only never stabilizes, RDBT bolts to beast-ears, structure tags fix it.
Anima is a Cosmos-based DiT, not SDXL, so one Anima LoRA loads on every derivative checkpoint. 20+ CivitAI Anima models sorted by type, aesthetic, and prompt adherence.
Tested on M1 Max ComfyUI: two WAI-Anima character LoRAs in one image. Side-by-side works, but only non-overlapping inpaint keeps the design pixel-exact; Qwen-Image-Edit elongates the side ponytail even with training tags, and overlapping interaction poses jam at skeleton extraction.
Tested on ComfyUI with Anima Turbo LoRA: hair intakes don't fire from the single tag. The 27-condition recipe — character NL reference + negative identity strip + (hair intakes:1.5) + 22.5° upper-front camera — also revealed the generic prompt was accidentally a Blue Archive character spec, and that Kanon's whole female cast has intakes as a cluster prior.
On M1 Max with Anima-Base v1.0 and WAI-Anima v1: the official negative is short. Trim the long Illustrious bad-hands list, move structure to positive tags, keep `safe` upfront.
Ran WAI-Anima v1.0 with a custom character LoRA on an M1 Mac to see if 2- and 3-character compositions actually hold up. Notes on what breaks and what holds at different LoRA weights, with practical settings that stay stable.
Tested on M1 Max 64GB: SetLatentNoiseMask returns unchanged images on Anima and Anima-Turbo with no error. LanPaint alone takes 32 min per image; adding Inpaint-CropAndStitch drops text inpaint to 2:31 and clothing replacement to about 7 min.
Tested on M1 Max 64GB ComfyUI: Anima-Base v1.0 matches preview3-base in speed; WAI-Anima kana LoRA hits 22% on light prompts but 67% with hood+robe+embroidery added.
LoRAs trained directly on Z-Image-Turbo come out blurry at 8 steps but fine at 30 — the distilled trajectory breaks. How Ostris's zimage_turbo_training_adapter avoids this as training-time-only scaffolding, plus SDXL LoRA incompatibility and Z-Image caption notes.
Three local image generation engines (WAI-Anima, WAI-IL/SDXL, FLUX.2 Klein 4B) tied together by a thin FastAPI wrapper that takes Japanese prompts. Ollama (gemma3:12b) handles JP→EN, ComfyUI workflows are built on the fly in Python, FLUX.2 runs as an mflux subprocess, and the whole thing is reachable from an iPhone over Tailscale.
