The two images from the manga mode showcase were taken from Say Hello to Black Jack, a manga in the public domain.
The results with these images are somewhat poor, but more recent manga typically yield better outcomes.
Original | MangaSD | MangaXL |
- Model: AOM3
- VAE: stabilityai mse-840000-ema
- Clip Skip:
2
ControlNet Configuration
- UNIT 1 : control_v11p_sd15s2_lineart_anime (YAML)
- UNIT 2 : control_v11p_sd15_softedge (YAML)
Inference Configuration
- Enable controlnet lowvram
- Keep original picture size
- Use interrogation
- Include metadata
- Use YoloV8 segmentation
- YoloV8 ONNX Model : manga_model
- Seed :
100
- Sampler : Euler
- Steps :
12
- Model: AbyssOrange XL Else
- VAE: None
- Clip Skip:
2
ControlNet Configuration
- UNIT 1 : MistoLine_rank256
- UNIT 2 : MistoLine_rank256
Inference Configuration
- Enable controlnet lowvram
- Keep original picture size
- Use interrogation
- Include metadata
- Use YoloV8 segmentation
- YoloV8 ONNX Model : manga_model
- Seed :
100
- Sampler : Euler
- Steps :
12
The two images from the drawing mode showcase are in the public domain.
Original | DrawingSD | DrawingXL |
- Model: AOM3
- VAE: stabilityai mse-840000-ema
- Clip Skip:
2
ControlNet Configuration
- UNIT 1 : control_v11p_sd15s2_lineart_anime (YAML)
- UNIT 2 : control_v11p_sd15_canny (YAML)
Inference Configuration
- Enable controlnet lowvram
- Keep original picture size
- Use interrogation
- Include metadata
- Use YoloV8 segmentation
- Seed :
100
- Sampler : Euler
- Steps :
12
- Model: AbyssOrange XL Else
- VAE: None
- Clip Skip:
2
ControlNet Configuration
- UNIT 1 : MistoLine_rank256
Inference Configuration
- Enable controlnet lowvram
- Keep original picture size
- Use interrogation
- Include metadata
- Use YoloV8 segmentation
- Seed :
100
- Sampler : Euler
- Steps :
12