Same Prompt, Same Seed: Anima vs SDXL Image Quality Compared
nano_chad99 · reddit · 2026-09-11
A Reddit user ran the exact same prompt, seed (188999578432144420), steps (25), sampler (Euler a), CFG (6) and size (1024x1280) through an Anima model (waiANIMAv10Base10 with qwen306bbase + qwenimagevae) and an SDXL model (waiIllustriousSDXLv160 with xlVAEC) to compare output quality.
The author reports Anima outputs are consistently lower quality: roughly 50% of generations (SFW or not) are usable and 50% are garbage, and even the good ones lag SDXL. Yet Civitai shows others getting high-quality Anima results, so he suspects a settings problem and asks for help.
Key differences: Anima used Schedule type Normal, Shift 3, RNG CPU; SDXL used Schedule type Automatic, Clip skip 2, plus a stack of negative TI embeddings (lazyneg, SmoothNegative, SmoothNegativeHands).
More from Multimodal
- Ethan Mollick prompts Google's Astra to build a game about the end of all things — eldonredwards · 2026-09-11
- Cristóbal Valenzuela unveils Solaris, an "interface world model" — c_valenzuelab · 2026-09-11
- World's first AI idol Yuri holds a concert with a stunning opening — xiaohu · 2026-09-11
- AI short film "The World Worth Living In" enters global AI film festivals — Xanavibes · 2026-09-11
- Training a 210M text-to-image DiT on one GPU in 3.5 days: three measured findings — IvanMikhnenkov · 2026-09-11
- Synthesia Adds Custom Avatars Built From a Text Prompt — synthesiaIO · 2026-09-11