Ripened版本v1.4 (ID: 624247)

Somewhat informative information

 

This model mix main focus is to create highly detailed anime-styled portraits, concept artworks as well as cinematic sceneries. It is decent at posing based on commands alone and should work with most SD1.5 LoRAs. The model has a fairly distinct style, so the threshold of applied LoRAs might have to be higher than described on their respective model cards. Make sure the model is loaded with clip skip 2 before generating images. VAE is baked in.

While this model produces good images as is, without any negative prompts required, below are listed a selection of positive as well as negative prompt tags and embeddings that I found to improve generation results. Unlike for the negatives, I do not recommend using all of the positive prompt examples together at once in case you are looking for a very specific outcome. You can also just copy and paste the negative prompt from the example images, given you downloaded the embeddings used.

Positive:

Silhouette Art, breathtaking manga, volumetric lighting, cinematic anime, skin texture, goosebumps, high contrast detail, abstract expressionist, award-winning anime artwork

Negative:

worst quality, low quality, normal quality, lowres, bad anatomy, bad hands, monochrome, grayscale, watermark, gloss

Recommended Embeddings:

• Aissist

• Beyond Negative

• Deep Negative v1.75t

• KNXO

• bad_pictures

• negativehand-negative-embedding

• EasyNegative

• veryBadImageNegative

• bad-hands-5

Knowing the comments – yes - this model is also capable of producing NSFW content, so if you are looking for that purpose solely, there are enough LoRAs on this platform to pair with.

Recommended upscalers:

4x-PBRify_UpscalerSIR-M_V2

4x_NMKD-Siax_200k

 

Recommended resolution:

512x768

768x512/768

I would personally not recommend going much higher than 768 as the results are not improving much.

 

Recommended generation settings (based on 512x768):

| Sampler               | Sampling Steps | CFG | Upscaling Steps | Upscale By | Den

|-----------------------|----------------|-----|-----------------|------------|-----

| DPM++ SDE Karras      | 15             | 6   | 12              | 1.5-2      | 0.5       

| DPM++ 3M SDE Karras   | 30             | 6   | 10              | 1.5-2      | 0.5       

| DPM++ 2S a Karras     | 20-40          | 7   | 8               | 2          | 0.4

| Restart               | 30             | 6   | 10              | 1.75-2     | 0.5         

Generation environment

All of the preview images were generated using Automatic1111 using an Rx 6800xt (ROCm toolkit enabled)

描述:

训练词语: