π΄ Dream Diffusion Pony - REVISED - By DICE π΄
NOTICE: REVISED VERSION
Was pointed out to me that I had uploaded the wrong file. This is now the correct file to download. Sorry about that guys. You will notice a dramactic increase in render quility with this version. ENJOY
Runs a nice natural 3D look. You can get awesome results without using LoRAs but try both ways see what differences it creates. (Use with causion as this Model will create NSFW content).
Suggested Uses For Prompts....
Always use: score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up
I Always start my prompts with.. (score_9, score_8_up, score_7_up), canon eos 5d mark iv, highly detailed, photorealistic, raw photo, depth of field, ..... Gets me that nice image between real and cartoon.
Use Clip Skip 2
VEA: sdxl_vae.safetensors is Baked in already
Certain keywords will have a bigger effect on the outputs, they are: Realistic, Dynamic poses, Highly Detailed, Attention to details, Canon EOS 5D Mark IV, UHD
When prompting I work a CFG between 6 to 9, depending on lengh of your prompts.
Because it's a Pony base, you can utilize the data selection tags of: 'source_pony', 'source_furry', 'source_cartoon' and 'source_anime' and ratings of 'rating_safe', 'rating_questionable' and 'rating_explicit'.
Other tips: Hires fix I use RealESRGAN_x4plus. I've also run a lot of renders using the sampler of DPM++2M Turbo.
Example of good prompt for this model is something along these lines:
Positive:
(score_9, score_8_up, score_7_up), canon eos 5d mark iv, highly detailed, photorealistic, raw photo, depth of field, 1woman, (UPSIDE-DOWN), (LOOKING THRU LEGS). hands on back of legs, (wearing a black Thong and white blouse, black hold ups, high heel shoes), Painstaking Attention To Details, UHD, <lora:Thru_Legs_-_By_Dice:1>
Negative:
blurry face, blurry eyes, poorly drawn, sketch, censored, lowres, low detail, blurry, watercolor, muscular female, bad anatomy, bad proportions, deformed, deformed anatomy, deformed fingers, too few fingers, too many fingers,
Full Workflow for Auto1111 and Forge:
(score_9, score_8_up, score_7_up), canon eos 5d mark iv, highly detailed, photorealistic, raw photo, depth of field, 1woman, Blonde, (UPSIDE-DOWN), (LOOKING THRU LEGS). hands on back of legs, (wearing a black Thong and white blouse, black hold ups, high heel shoes), Painstaking Attention To Details, UHD, <lora:Thru_Legs_-_By_Dice:1>
Negative prompt: blurry face, blurry eyes, poorly drawn, sketch, censored, lowres, low detail, blurry, watercolor, muscular female, bad anatomy, bad proportions, deformed, deformed anatomy, deformed fingers, too few fingers, too many fingers,
Steps: 36, Sampler: DPM++ 2M Turbo, CFG scale: 8, Seed: 246483945, Size: 544x824, Model hash: cac44c72fd, Model: Dream_Diffusion_Pony_REVISED_By_DICE, VAE hash: 8db2075341, VAE: sdxl_vae.safetensors, Clip skip: 2, Hires upscale: 2, Hires steps: 20, Hires upscaler: RealESRGAN_x4plus, Lora hashes: "Thru_Legs_-_By_Dice: 5c784ba0e5e1", Version: f0.0.17v1.8.0rc-latest-276-g29be1da7
Outcome Render:
Description
Dream Diffusion Pony - By DICE
NOTICE: REVISED VERSION
Was pointed out to me that I had uploaded the wrong file. This is now the correct file to download. Sorry about that guys. You will notice a dramactic increase in render quility with this version. ENJOY
Runs a nice natural 3D look. You can get awesome results without using LoRAs but try both ways see what differences it creates. (Use with causion as this Model will create NSFW content).
Suggested Uses For Prompts....
Always use: score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up
I Always start my prompts with.. (score_9, score_8_up, score_7_up), canon eos 5d mark iv, highly detailed, photorealistic, raw photo, depth of field, Gets me that nice image between real and cartoon.
Use Clip Skip 2
VEA: sdxl_vae.safetensors is Baked in already
Certain keywords will have a bigger effect on the outputs, they are: Realistic, Dynamic poses, Highly Detailed, Attention to details, Canon EOS 5D Mark IV, UHD
When prompting I work a CFG between 6 to 9, depending on lengh of your prompts.
Because it's a Pony base, you can utilize the data selection tags of: 'source_pony', 'source_furry', 'source_cartoon' and 'source_anime' and ratings of 'rating_safe', 'rating_questionable' and 'rating_explicit'.
Other tips: Hires fix I use RealESRGAN_x4plus. I've also run a lot of renders using the sampler of DPM++2M Turbo.
Example of good prompt for this model is something along these lines:
Positive:
(score_9, score_8_up, score_7_up), canon eos 5d mark iv, highly detailed, photorealistic, raw photo, depth of field, 1woman, (UPSIDE-DOWN), (LOOKING THRU LEGS). hands on back of legs, (wearing a black Thong and white blouse, black hold ups, high heel shoes), Painstaking Attention To Details, UHD, <lora:Thru_Legs_-_By_Dice:1>
Negative:
blurry face, blurry eyes, poorly drawn, sketch, censored, lowres, low detail, blurry, watercolor, muscular female, bad anatomy, bad proportions, deformed, deformed anatomy, deformed fingers, too few fingers, too many fingers,
Full Workflow for Auto1111 and Forge:
(score_9, score_8_up, score_7_up), canon eos 5d mark iv, highly detailed, photorealistic, raw photo, depth of field, 1woman, Blonde, (UPSIDE-DOWN), (LOOKING THRU LEGS). hands on back of legs, (wearing a black Thong and white blouse, black hold ups, high heel shoes), Painstaking Attention To Details, UHD, <lora:Thru_Legs_-_By_Dice:1>
Negative prompt: blurry face, blurry eyes, poorly drawn, sketch, censored, lowres, low detail, blurry, watercolor, muscular female, bad anatomy, bad proportions, deformed, deformed anatomy, deformed fingers, too few fingers, too many fingers,
Steps: 36, Sampler: DPM++ 2M Turbo, CFG scale: 8, Seed: 246483945, Size: 544x824, Model hash: cac44c72fd, Model: Dream_Diffusion_Pony_By_DICE, VAE hash: 8db2075341, VAE: sdxl_vae.safetensors, Denoising strength: 0.65, Clip skip: 2, Hires upscale: 2, Hires steps: 20, Hires upscaler: RealESRGAN_x4plus, Lora hashes: "Thru_Legs_-_By_Dice: 5c784ba0e5e1", Version: f0.0.17v1.8.0rc-latest-276-g29be1da7
Outcome Render: