A streamlined creative image workflow
5.0
0 reviewsDescription
The workflow, integrating both LLAMA and style plugins, is notable for its utilization of CFG scale compression and conditional proportionality, affording greater control over textual influence on the sampler during image generation.
To optimize CFG compression, consider employing triangular CFG scaling, initiating at a value of 1 or 1.5. This facilitates setting the sampler's CFG value to 7 or higher (particularly beneficial when utilizing TCD samplers for enhanced image control). Additionally, leveraging conditional averaging allows for nuanced control over the influence of style prompts versus LLAMA prompts. A lower value amplifies the impact of style prompts, while a higher value prioritizes LLAMA's guidance in image generation, mitigating potential counteraction from LLAMA.
Regarding LORA, maintain the default settings for TCD LORA. For detail-enhancing LORA, consider adjusting values within the 4-10 range. This amplifies the influence of CFG on the generated image.
Discussion
(No comments yet)
Loading...
Reviews
No reviews yet
Versions (1)
- latest (2 years ago)
Node Details
Primitive Nodes (2)
MZ_LLama3CLIPTextEncode (1)
PrimitiveNode (1)
Custom Nodes (20)
- DeepTranslatorTextNode (1)
- CR Apply LoRA Stack (1)
ComfyUI
- FreeU_V2 (1)
- CLIPTextEncode (2)
- VAEDecode (1)
- CLIPSetLastLayer (1)
- CheckpointLoaderSimple (1)
- EmptyLatentImage (1)
- SaveImage (1)
- VideoTriangleCFGGuidance (1)
- ConditioningAverage (1)
- LatentRotate (1)
- SamplerCustom (1)
- easy loraStack (1)
- LayerFilter: HDREffects (1)
- SDXLPromptStylerPreview (1)
- TCDModelSamplingDiscrete (2)
- ShowText|pysssss (1)
Model Details
Checkpoints (1)
juggernautXL_v8Rundiffusion.safetensors
LoRAs (5)
MJ模仿\add_detail.safetensors
常用\TCD-LORA.safetensors
常用\鲜创一派_外星访问XL_V1.0.safetensors
常用\鲜创一派_幻想生物XL_V1.0.safetensors
新下\哑光质感- SDXL - vanta-black_contrast_V3.0.safetensors