zavymix

For best outputs:

• Version 2.0:

  • Sampler: DPM++ 2M Karras.

  • Steps: 30, cfg 7, CLIP skip 2.

  • VAE: kl-f8-anime2_fp16.safetensors

  • Pos prompt: add the LoRa's "more_detail" with a weight of 0.8, "epiNoiseoffset_v2-pynoise" with a weight of 1.5 and "LowRA" with a weight of 0.3 at the end of the prompt (included in the v2.0 example run).

  • Neg prompt: too long to be pasted in the readme, but feel free to use the included in both versions' example runs. Make sure you include the embeddings "bad-hands-5" and "bad_prompt_version2-neg" (included in both versions' example runs).

• Version 1.0:

  • Sampler: DPM++ 2M Karras.

  • Steps: 40, cfg 8.5, CLIP skip 2.

  • VAE: clearvae_v2.3_fp16.safetensors

  • Pos prompt: add the LoRa's "add_detail" with a weight of 0.4 and "epiNoiseoffset_v2-pynoise" with a weight of 1.5 at the end of the prompt (included in the v1.0 example runs).

  • Neg prompt: too long to be pasted in the readme, but feel free to use the included in both versions' example runs. Make sure you include the embeddings "bad-hands-5" and "bad_prompt_version2-neg" (included in both versions' example runs).

• What changes from v1.0 to v2.0: input_image feature, overall quality increased.

Checkpoint Merge.