/

/

anima yume

anima yume

AD

About This Model

ANIMA · Base Models

(Updated to support Anima LoRAs!) The "AnimaYume" fine-tune (by duongve13112002) of Anima Base (v1) created by CircleStone Labs. It has a "Turbo" mode and a "Normal" (default) mode. This Workflow features optionally features a Turbo LoRA as well (also by CircleStone Labs) and is activated when using a CFG / Guidance of 1 (the default guidance is 3.0 if not set, meaning this lora won't be used). Note that this tends to have its own default style, unless you guide it elsewhere via your prompt. The left of the example image showcases "normal" vs "turbo" which is on the right. When using the "normal" mode (any CFG that isn't 1), a negative prompt is heavily recommended, even something like: "blurry, low quality, bad quality, censor bar, jpeg artifacts, sepia, ugly, deformed, watermark, logo, signature" As is usually the case, setting a CFG of 1 (to activate the turbo lora) will disable negative prompts. Adjusting step count is supported, however in the default normal mode, steps will always at minimum be 20, whereas the minimum for turbo mode is 8 steps. The workflow will toggle between 20 (normal) vs 8 steps (turbo) if left to the default. slot1 can also be used to adjust the sampler, using the following values below: 0 - er_sde 1 - euler_ancestral 2 - dpmpp_2m_sde_gpu 3 - res_2m (Default) 4 - res_multistep You can also use `/scale` to perform an automatic Tiled Hi-Res Fix upscale of the image, such as `/scale:2.0` - the `/strength` flag changes the "denoise" value of the upscal