Loading
Tem um workflow do ComfyUI? Gere uma documentação legível direto no navegador.Torne um workflow legível.Experimentar agora ↗
Originally posted to HuggingFace by Stkzzzz222 and shared on Reddit .
Fonte do modelo
Descrição da fonte
Originally posted to HuggingFace by Stkzzzz222 and shared on Reddit .
Disclaimer: Remix is very experimental, it's possible to don't obtain good results, maybe changing the order of the elements that you want to mix or giving them more weight can help. Its use will be under your own risk.
Remix can be used in 2 ways.
As an artist using (prompt) "by remix"
Comparisions for the same prompt and settings without and with ",by remix"
Examples:
a cow with robotic legs. by remix Negative prompt: ((overexposure)), ((high contrast)),(((cropped))), (((watermark))), ((logo)), ((barcode)), ((UI)), ((signature)), ((text)), ((label)), ((error)), ((title)), stickers, markings, speech bubbles, lines, cropped, lowres, low quality, artifacts Steps: 20, Sampler: DPM++ SDE Karras, CFG scale: 7, Seed: 362581665, Size: 768x768, Model hash: 2c02b20a, ENSD: -1
Fontes
1 fonteArtefatos de modelo
1 artefatoTrechos de fonte
1 trechoSource context: 825 downloads · Type TextualInversion · Base model SD 2.0 768
a bouquet of planets by remix Negative prompt: ((overexposure)), ((high contrast)),(((cropped))), (((watermark))), ((logo)), ((barcode)), ((UI)), ((signature)), ((text)), ((label)), ((error)), ((title)), stickers, markings, speech bubbles, lines, cropped, lowres, low quality, artifacts Steps: 28, Sampler: Euler a, CFG scale: 8, Seed: 1458414194, Size: 768x768, Model hash: 2c02b20a, ENSD: -1
a bouquet of aliens by remix Negative prompt: ((overexposure)), ((high contrast)),(((cropped))), (((watermark))), ((logo)), ((barcode)), ((UI)), ((signature)), ((text)), ((label)), ((error)), ((title)), stickers, markings, speech bubbles, lines, cropped, lowres, low quality, artifacts Steps: 28, Sampler: Euler a, CFG scale: 8, Seed: 532558973, Size: 768x768, Model hash: 2c02b20a, ENSD: -1
As a mixer of other elements using "by remix" at the end to reinforce the result
a (element A) remix a (element B). very detailed, intrincated. by remix
Examples:
a pig remix a ladybug. very detailed, intrincated. by remix Negative prompt: ((overexposure)), ((high contrast)),(((cropped))), (((watermark))), ((logo)), ((barcode)), ((UI)), ((signature)), ((text)), ((label)), ((error)), ((title)), stickers, markings, speech bubbles, lines, cropped, lowres, low quality, artifacts Steps: 28, Sampler: Euler a, CFG scale: 8, Seed: 762655912, Size: 768x768, Model hash: 2c02b20a, ENSD: -1
a cow remix a fish. very detailed, intrincated. by remix Negative prompt: ((overexposure)), ((high contrast)),(((cropped))), (((watermark))), ((logo)), ((barcode)), ((UI)), ((signature)), ((text)), ((label)), ((error)), ((title)), stickers, markings, speech bubbles, lines, cropped, lowres, low quality, artifacts Steps: 28, Sampler: Euler a, CFG scale: 8, Seed: 1458414194, Size: 768x768, Model hash: 2c02b20a, ENSD: -1
a dog remix spaceship, very detailed, intrincated by remix Negative prompt: ((overexposure)), ((high contrast)),(((cropped))), (((watermark))), ((logo)), ((barcode)), ((UI)), ((signature)), ((text)), ((label)), ((error)), ((title)), stickers, markings, speech bubbles, lines, cropped, lowres, low quality, artifacts Steps: 28, Sampler: Euler a, CFG scale: 8, Seed: 555988660, Size: 768x768, Model hash: 2c02b20a, ENSD: -1
These are the basic ways of using it, but it can also be used with other kind of prompts, experimentation is encorauged.
For example:
a giraffe AND spider, very detailed, intrincated, by remix Negative prompt: ((overexposure)), ((high contrast)),(((cropped))), (((watermark))), ((logo)), ((barcode)), ((UI)), ((signature)), ((text)), ((label)), ((error)), ((title)), stickers, markings, speech bubbles, lines, cropped, lowres, low quality, artifacts Steps: 28, Sampler: Euler a, CFG scale: 8, Seed: 3615455169, Size: 768x768, Model hash: 2c02b20a, ENSD: -1
Remix also has his own avatar
If you write Remix as a prompt there are many possibilities that he appears. It's strange because in the dataset there weren't any human figures, and his apperance has been changing as the embedding was trainned. So if he appears in your experiments, say hello to him.
