Version V8 differs from previous versions in that the large model has been retrained with the latest GPT-4 low-quality annotation and Cog natural language complementary details. Hand fine-tuning, photoreflections, and color have also been retrained into this retrained large-scale model.

Recommended iteration steps: 25-30 steps.

Recommended sampling method: DPM++ 2M Karras

Native webui 1.9.0 name: DPM++ 2M

Recommended amplification algorithm: R-ESRGAN_4x+ Anime6B

Enable After Detailer when needed: face_yolov8n_v2.pt

This model has been improved to enhance the dark effect, and the light and glare have been carefully adjusted. Compared with the previous version, the overall texture has been significantly improved.

recommend iteration steps: 25

Sampling method for reference recommend:(DPM 2M Karras)

recommend amplification algorithm: R-ESRGAN_4x Anime6B

Turn on AD when needed

[Advanced official version, incorporating colorful beauty pupil effect, color depth, drawing details, clarity, live animation.]

Advanced official version adjustment parameters:

Sampling method for reference recommend:(DPM 2M Karras)(DPM SDE Karras)(DPM2)(Euler a)

recommend iteration steps: 30-45,

Recommendations for redrawing amplitude: (default 7)

Size recommendations: 768, 1024,2048

Suggestion of prompt words: (default recommendation 10 coefficient)