juggernautXL

CHECKPOINT

转载

LQ兰雀

Mar 25, 2024 6:22 AM 更新

A sad man standing in front of a burnt down home in a forest, stunning details, 8k

beautiful lady, (freckles), big smile, ruby eyes, long curly hair, dark makeup, hyperdetailed photography, soft light, head and shoulders portrait, cover

Leica portrait of a gremlin skateboarding, coded patterns, sparse and simple, uhd image, urbancore, sovietwave, period snapshot

Leica Hasselblad portrait, hyperdetailed Photography, a Native American man walks proudly confidently in traditional clothing with an Indian feather hat on his head a wolf accompanies him in the streets of New York in the middle of the buildings in the middle of the road. Large view

A hyperdetailed photograph of a Cat dressed as a mafia boss holding a fish walking down a Japanese fish market with an angry face, 8k resolution, best quality, beautiful photograph, dynamic lighting,

A beautiful portrait photograph of a dragon with diamond and gemstone scales, opal eyes, cinematic, gem, diamond, crystal, fantasy art, hyperdetailed photograph, shiny scales, 8k resolution,

Michael Jackson singing at a standing microphone on a super-wide stage in a really big arena, by conrad roset, greg rutkowski, makoto shinkai, low poly, abstract, style of black and white drawing with red pops of color

An evil demon crawling out of a mirror with a horrifying reflection on a polished wood floor, mirror has a ((silver frame)), 8k resolution, best quality, hyperdetailed photograph, lowlight

moody aesthetic, beautiful cozy, cramped bedroom with floor to ceiling glass windows overlooking a cyberpunk city at night, view from top of skyscraper, white bedsheets, bookshelves, thunderstorm outside with torrential rain, detailed, high resolution, photorrealistic, dark, gloomy,

a beautiful cottage, scotland coastal village, (sharp focus:1.2), extremely detailed, (photorealistic:1.4), (RAW image, 8k high resolution:1.2), RAW candid cinema, 16mm, color graded Portra 400 film, ultra realistic, cinematic film still, subsurface scattering, ray tracing, (volumetric lighting)

photograph, landscape of a Mythical Grotto from inside of a Harare, at Twilight, Depressing, Cloudpunk, Cold Lighting, dynamic, Nikon d850, Depth of field 270mm, Amaro, Golden ratio

A woman wearing a hat poses for a picture, in the style of oshare kei, black, wide lens, shiny/ glossy, solapunk, dark silver, rim light

Portrait of a kind of a young man, twenty years with brown hair, childish, droopy eyes, unconventional beauty, moody atmosphere by Saul Leiter

Medium shot, Adorable creature with big reflective eyes, moody lighting, best quality, full body portrait, real picture, intricate details, depth of field, in a forest, fujifilm xt3, outdoors, bright day, beautiful lighting, raw photo, 8k uhd, film grain, unreal engine 5, ray tracing

Cyberpunk, sci-fi, dark-fantasy, kodak portrait 400, 8k, soft light, volumetric lighting, highly detailed, britt marling style 3/4, portrait photo of a war man cyborg robot in a chemical laboratory + face, face is skullbone with scares, intricate, elegant, highly detailed, devil-armor, 2D motifs detailed dark fantasy digital painting, artstation, concept art, smooth, sharp focus, illustration, art by Otomo Katsuhiro and ShirÅ Masamune and Oshii Mamoru. Cosmic light in backfront

photograph, a path in the woods with leaves and the sun shining , by Julian Allen, dramatic autumn landscape, ears, park, take off, peace, rich cold moody colours, hi resolution, oaks

Coqui frog in batman uniform, 8k

A sad man standing in front of a burnt down home in a forest, stunning details, 8k

For business inquires, commercial licensing, custom models, and consultation contact me under juggernaut@rundiffusion.com

Juggernaut is available on the new Auto1111 Forge on RunDiffusion

JuggernautXL V9 + RunDiffusion Photo 2 on Hugging Face

A big thanks for Version 9 goes to RunDiffusion (Photo Model) and Adam, who diligently helped me test :) (Leave some love for them ;) )

It's time for another round, this time a bit delayed, but I hope you forgive the delay. Let's dive straight into the changes that await you or what we've been working on lately:

For V9, I myself have only done basic training. This involves some work on skin details, lighting, and overall contrast. However, the biggest change to the model came from the RunDiffusion Photo Model update, which was made available to me in V2 by RunDiffusion. The photographic output of the model should, in our experience, be even stronger than in previous versions.

Now for a small "roadmap" update, or a general status update on how things are progressing with Juggernaut. As you may have noticed, there was a slight delay with V9. With each successive version, it has become increasingly difficult to train Juggernaut without sacrificing quality in some areas, which was already the case to some extent with V8. Don't worry, V9 is really good, and I'm satisfied with the version I can present to you today :) However, I've decided to go for a complete "reboot" for V10. I want to simply retrain the Juggernaut base set. The conditions for better captioning weren't as favorable "back then" as they are today, so I want to completely re-caption the base set (5k images) with GPT-4 Vision. I expect a big leap towards prompting guidance and quality.

But as you surely noticed last week, the release of Stable Cascade got in the way a bit. Therefore, my focus in the coming weeks will be on training Juggernaut for Stable Cascade. The approach remains the same as with the planned "reboot"; I want to caption/tag all images in the future only with GPT-4 or manually. The timeline for all of this is still uncertain. I hope to be able to present you with a first stable version of Juggernaut Cascade sometime in March. V10 of Juggernaut XL will follow in the weeks thereafter.

Now, here are some additional tips to make prompting easier for you:

Res: 832*1216

Sampler: DPM++ 2M Karras

Steps: 30-40

CFG: 3-7 (less is a bit more realistic)

Negative: Start with no negative, and add afterwards the Stuff you don´t wanna see in that image. I don´t recommend using my Negative Prompt, i simply use it because i am lazy :D

VAE is already Baked In

HiRes: 4xNMKD-Siax_200k with 15 Steps and 0.3 Denoise + 1.5 Upscale

And a few keywords/tokens that I regularly use in training, which might help you achieve the optimal result from the version: