Common Sense Made Simple | What is CFG Scale in Stable Diffusion Automatic1111 img2img & Deforum Colab Notebooks @iamYork_ | Uploaded 1 year ago | Updated 7 hours ago
The classifier-free guidance scale (CFG Scale), another setting in the Stable Diffusion model, is similar to Seed. The CFG scale changes how much the image resembles the prompt or other input image. The output will be more in line with the input prompt and/or input image if CFG Scale is greater, but it will be distorted. On the other hand, the lower the CFG Scale value, the better the quality but the more likely it is to stray from the prompt or the input image... Here is my explanation of how I use batch img2img to achieve the results I want for the video input type.https://youtu.be/II064QvKwqo
Everything in the comparisons above is raw, straight from Automatic1111, and the only thing I did in post-production to help reduce the jitter was to lower the frame rate.... I normally do a bit of post work in Adobe After Effects here are some of the basics I use on that front if you use Ae: youtu.be/8w2qwIEd1KA
The fidelity between the prompt and output images and the value of the CFG scale are directly proportional to one another. The quality of the output and the value of the CFG Scale are inversely related... As the setting maxes out at 30 then it sort of goes Super Saiyan Saturated as Enigmatic_E likes to use... He loves his anime and excessive color saturation... To each their own I say I say I say AKA I reckon...
If you like what I am doing and would like to support me directly via paypal:
paypal.com/paypalme/iamYork667
We appreciate any support as this channel is 100% funded by independent artists and channel memberships donating their time and patience to making everything you see and enjoy on this channel...
JOIN this CHANNEL to get Access to CUSTOM Trained AI models, My Colab Notebooks, One on One AI Technical Support, Project Files, Settings, Copyright free Music and UNIQUE Text Prompts:
youtube.com/channel/UC11G7rbLqN6dtSHcpKNnIoQ/join
iamYork has also launched a limited clothing line where ALL PROFITS
go to independent artists: PLEASE INDULGE : etsy.com/shop/iamYork
iamYork Imagery is a boutique Design, Animation, and Visual Effects studio in Hell's kitchen, New York city created by Nicolas Fenrir Tengri, a mixed media artist, animator and anti-designer...
iamYork's Website iamyork.com
Royalty FREE No Copyright Music: pond5.com/artist/iamyork
Instagram: instagram.com/i.am.york
Facebook: facebook.com/ARTorFACT
Twitter: twitter.com/FenrirTengri
TikTok: tiktok.com/@i.am.york
NFT Art & Animation : Opensea: opensea.io/iamYork
ALL my tutorials and comparisons: youtube.com/playlist?list=PLVWl_nVCEeEzVmb0--9AEGHTL1pa97FGn
MORE Stable Diffusion Tutorials and Examples: youtube.com/playlist?list=PLVWl_nVCEeEwy_J_lrZOrelwF22Ikrbsp
If you just want to see more "AI generated" Anime this is that playlist: youtube.com/playlist?list=PLVWl_nVCEeExAXO8a18VjbZOHEFKcisW0
After Effects & Cinema 4D Tutorials: youtube.com/playlist?list=PLVWl_nVCEeEzveVd8tyXzOlSuP0-1ecav
My "traditional" Digital Animation and Independent Experimental Film Work can be found here: youtube.com/playlist?list=PLVWl_nVCEeEzujp534UtDOzOEcbFrA2k2
Settings for SD animation: I used Automatic1111's free web UI
Positive Text Prompt: Laughing Hyena Highly Detailed impasto oil painting, abstract expressionism, 4k, 8k
Negative Text Prompt: Depth of field, text, error, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, artist name
Steps: 67
CFG Scale: [variable]
Width: 704
Height: 712
Sampler: Euler ancestral
Original 3D animation and modeling all done using Cinema 4D...
With just a short and straightforward set of img2img instructions, the LAION dataset, on which the Stable Diffusion was trained, is capable of producing useful results. However, you might need to increase the CFG scale value and/or denoising strength if you're trying to generate something that Stable Diffusion is unfamiliar with or if you want to combine multiple concepts or people. Stable Diffusion will therefore produce imaginatively at the expense of image quality...
Stable diffusion AI: What is it?
A text-to-image model using machine learning called Stable Diffusion creates digital images from descriptions in natural language. The underlying methodology was created at LMU Munich and later expanded through a partnership between StabilityAI, LMU, and Runway with assistance from EleutherAI and LAION....
Stable Diffusion by Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, Björn Ommer and the Stability.ai Team.
K Diffusion by Katherine Crowson. and HuggingFace.
Chapters : Time Stamps:
0:00 - Original Render:
00:13 - 1.0:
00:27 - 3.0:
00:40 - 7.0:
00:54 - 10.0:
1:07 - 13.0:
1:21 - 15.0:
1:35 - 20.0:
1:48 - 25.0:
2:02 - 30.0:
2:15 - Brady Bunch GRID:
#stablediffusion #ai #comparison
The classifier-free guidance scale (CFG Scale), another setting in the Stable Diffusion model, is similar to Seed. The CFG scale changes how much the image resembles the prompt or other input image. The output will be more in line with the input prompt and/or input image if CFG Scale is greater, but it will be distorted. On the other hand, the lower the CFG Scale value, the better the quality but the more likely it is to stray from the prompt or the input image... Here is my explanation of how I use batch img2img to achieve the results I want for the video input type.https://youtu.be/II064QvKwqo
Everything in the comparisons above is raw, straight from Automatic1111, and the only thing I did in post-production to help reduce the jitter was to lower the frame rate.... I normally do a bit of post work in Adobe After Effects here are some of the basics I use on that front if you use Ae: youtu.be/8w2qwIEd1KA
The fidelity between the prompt and output images and the value of the CFG scale are directly proportional to one another. The quality of the output and the value of the CFG Scale are inversely related... As the setting maxes out at 30 then it sort of goes Super Saiyan Saturated as Enigmatic_E likes to use... He loves his anime and excessive color saturation... To each their own I say I say I say AKA I reckon...
If you like what I am doing and would like to support me directly via paypal:
paypal.com/paypalme/iamYork667
We appreciate any support as this channel is 100% funded by independent artists and channel memberships donating their time and patience to making everything you see and enjoy on this channel...
JOIN this CHANNEL to get Access to CUSTOM Trained AI models, My Colab Notebooks, One on One AI Technical Support, Project Files, Settings, Copyright free Music and UNIQUE Text Prompts:
youtube.com/channel/UC11G7rbLqN6dtSHcpKNnIoQ/join
iamYork has also launched a limited clothing line where ALL PROFITS
go to independent artists: PLEASE INDULGE : etsy.com/shop/iamYork
iamYork Imagery is a boutique Design, Animation, and Visual Effects studio in Hell's kitchen, New York city created by Nicolas Fenrir Tengri, a mixed media artist, animator and anti-designer...
iamYork's Website iamyork.com
Royalty FREE No Copyright Music: pond5.com/artist/iamyork
Instagram: instagram.com/i.am.york
Facebook: facebook.com/ARTorFACT
Twitter: twitter.com/FenrirTengri
TikTok: tiktok.com/@i.am.york
NFT Art & Animation : Opensea: opensea.io/iamYork
ALL my tutorials and comparisons: youtube.com/playlist?list=PLVWl_nVCEeEzVmb0--9AEGHTL1pa97FGn
MORE Stable Diffusion Tutorials and Examples: youtube.com/playlist?list=PLVWl_nVCEeEwy_J_lrZOrelwF22Ikrbsp
If you just want to see more "AI generated" Anime this is that playlist: youtube.com/playlist?list=PLVWl_nVCEeExAXO8a18VjbZOHEFKcisW0
After Effects & Cinema 4D Tutorials: youtube.com/playlist?list=PLVWl_nVCEeEzveVd8tyXzOlSuP0-1ecav
My "traditional" Digital Animation and Independent Experimental Film Work can be found here: youtube.com/playlist?list=PLVWl_nVCEeEzujp534UtDOzOEcbFrA2k2
Settings for SD animation: I used Automatic1111's free web UI
Positive Text Prompt: Laughing Hyena Highly Detailed impasto oil painting, abstract expressionism, 4k, 8k
Negative Text Prompt: Depth of field, text, error, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, artist name
Steps: 67
CFG Scale: [variable]
Width: 704
Height: 712
Sampler: Euler ancestral
Original 3D animation and modeling all done using Cinema 4D...
With just a short and straightforward set of img2img instructions, the LAION dataset, on which the Stable Diffusion was trained, is capable of producing useful results. However, you might need to increase the CFG scale value and/or denoising strength if you're trying to generate something that Stable Diffusion is unfamiliar with or if you want to combine multiple concepts or people. Stable Diffusion will therefore produce imaginatively at the expense of image quality...
Stable diffusion AI: What is it?
A text-to-image model using machine learning called Stable Diffusion creates digital images from descriptions in natural language. The underlying methodology was created at LMU Munich and later expanded through a partnership between StabilityAI, LMU, and Runway with assistance from EleutherAI and LAION....
Stable Diffusion by Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, Björn Ommer and the Stability.ai Team.
K Diffusion by Katherine Crowson. and HuggingFace.
Chapters : Time Stamps:
0:00 - Original Render:
00:13 - 1.0:
00:27 - 3.0:
00:40 - 7.0:
00:54 - 10.0:
1:07 - 13.0:
1:21 - 15.0:
1:35 - 20.0:
1:48 - 25.0:
2:02 - 30.0:
2:15 - Brady Bunch GRID:
#stablediffusion #ai #comparison
