Counterfeit is a female focused high-quality anime model that can create stylized images from close-up portraits to long-distance shots. The model has a user-friendly prompt language that supports Danbooru tags, but "EasyNegative" embedding is essential for high-quality outputs. Additionally, it can work at different resolutions compared to other models.
What This Model Excels At
- Generating female/male images.
- Generating SFW/NSFW content.
- Crafting highly creative compositions.
- Crafting stylized anime style arts.
- Danbooru tags can be used while prompting, natural language prompts can be effective as well.
- Works well with various resolutions.
- Generating abstract/landscape illustrations.
Cautions For Optimal Usage
- Highly towards to generate feminine/female people.
- Using "EasyNegative v1 or v2" embedding can increase quality of image to highest levels, highly recommended to use.
- There is no clear difference between embedding versions, "v2" is trained on "Counterfeit v3" model. User choose one of them to use, model doesn't require any other negative in most of the situations.
- There is higher possibility to get anatomical errors because model prioritize the freedom of composition. (Very bad hand generations)
- Can't recognize most of well-known characters, only generate with some of Danbooru character tags. e.g. `mona \(genshin impact\)`
Recommended Settings
- Steps: 20-25
- CFG Scale: 8-10
- Sampler: DPM++ 2M Karras, DPM++ SDE Karras
- Clip Skip: 2
- VAE: None, kl-f8-anime2
- Upscale: R-ESRGAN 4x+ Anime6B
- Upscale Sampling Times: 10-15
- Redraw Noise Intensity: 0.6
Suggested Resources
Embedding
Prompt Style
Can work with both of tag based and natural language prompts.
- "EasyNegative" or "EasyNegativeV2" trigger words must be used on negative prompt, if user going to use embeddings. Don't use them together and pick one version to use. They don't have any significant difference but this is a recommendation for embedding's training:
EasyNegative: Counterfeit v2/v2.5
EasyNegativeV2: Counterfeit v3
- Works well with quality modifier prompts like `masterpiece, best quality, highly detailed, absurdres` etc.
- Great usage of `depth of field`
- Recognizes most of Danbooru tags.
- Does not need any other negative prompt with embedding, but they still can be used.
- `girl` is a trigger word for model. This means model will generate female focused images, if you use `girl` anywhere in your prompt.
Recommended Prompt Order
- Quality/Detail Terms
- Angle/View
- Subject
- Subject Details
- Additional Terms/Tags
- Lighting/Shadow
Prompt Examples
Office Senior
Positive Prompt
`(masterpiece, best quality), highly detailed, photorealistic artwork, sfw, (from above), 1girl, brown eyes, glasses, blush, green hair, hair ornament, big breasts, office lady, formal, standing, looking up, looking at viewer, indoors, office background, office desk, office chair, (depth of field), (bokeh:1.4), (pastel colors:0.8), [film grain], (by ligne claire:1.2), absurdres, dramatic studio lighting, soft shadows`
Negative Prompt
`EasyNegativeV2, nsfw, child, childish`
Date At Night
Positive Prompt
`(masterpiece, best quality:1.3), sfw, solo, dutch angle, 1girl, red eyes, serious, frown, black hair, long hair, straight hair, crop top, midriff, cleavage, leather jacket, black jacket, black pants, outdoors, street background, (lamppost:1.2), (night), sharp focus, depth of field, (subsurface light scattering), backlighting, absurdres`
Negative Prompt
`EasyNegativeV2, nsfw, child, childish`
You can join SeaArt Discord Community for more content, also you will have opportunity to choose next model guides with your vote. You can use comments or SeaArt server #guides-feedback to leave feedback or suggestion. We look forward to seeing you on the server.
Happy Generations To Everyone!