Custom Lora's and commissions here: https://ko-fi.com/suppressor
Updating for Wan! All Wan models are compatible with 2.1 and 2.2.
Hey all, currently updating this model by training individual Disney Princesses for WAN T2V 14B. I hope you enjoy. Will release each princess as the training finishes.
This is a Hunyuan video lora designed to create the likeness of Dis*ey Princesses from "Ralph Breaks the Internet," with some stylistic differences and choices mixed in from freely available OG Dis*ey cartoons and standardization images from the darkest dankest corners of the internet. Depending on strength you can get more or less CGI. The higher the strength, more D*sney, the lower, more realism.
Example Prompt Trigger: CGI Disney Princess Anna, with her auburn hair and twin braids is wearing her signature green top and shorts. Anna has large expressive eyes and blushed cheeks. Anna is smiling at the camera and performing a comedy routine.
Supported Characters that have been trained:
Anna
Ariel
Aurora
Belle
Cinderella
Elsa
Jasmine
Merida
Moana
Namaari
Pocahontas
Snow White
Tiana
Other D*sney Girls in the Mix (trained with minimal images and can be hit or miss)
Alice
Asha (sometimes)
Chel
Dolores
Esmerelda
Isabela
Jane
Kida (sometimes)
Megara (sometimes)
Mirabel
Raya (sometimes)
Shank (sometimes)
Venellope (sometimes)
This ALL DP version does a few things:
1) It allows 1 lora for most if not all D*sney Princesses and supporting cast.
I've had success prompting for multiple DP's in several styles, but you need clearly delineate in your prompt what you are trying to accomplish. Crucial items like outfit, hairstyle, skin tone and ethnicity must be prompted for.
2) It allows you to create your own Disney Princess if you can prompt well.
I've come up with some total babes that I will share soon. You can just create your ideal princess, crank up the strength, and let it ride. It works surprisingly well.
3) The strength directly correlates with the CGI and Cartoon effects.
4) It directly supports nudity and NSFW
Single Models:
RAPUNZEL
Sample prompt: Rapunzel, with her insanely long blonde hair is sitting at a bar on a chair wearing her purple dress. She is obviously drunk and seems a bit disoriented, her expression is both surprised and curious.
You can get more realism by decreasing the strength, keep it high for CGI.
BELL - BEAUTY AND THE BEAST
Pretty evenly balanced, just use the trigger and prompt for styles and expressions.
Sample prompt:
in a luxurious medieval ballroom, full body shot of white girl cartoon CGI Belle, dancing , alluring dance moves. the video is a computer-generated 3D render and Disney stylized. She is depicted in a cartoonish, anime-inspired style with exaggerated features big expressive green eyes, and vibrant colors. The character has a slender, build, her brown hair is in an updo. She is wearing a sleeveless, light yellow dress. She is flirting with the camera, being sexy, looking at the camera as if she is confused or curious starting to smile!
Add detailed texture work, dynamic shadows and godrays, bokeh, and lifelike animations to showcase the highest quality of CGI rendering, Chiaroscuro lighting,
AURORA - SLEEPING BEAUTY
Pretty evenly balanced. Always prompt Aurora Sleeping Beauty for CGI style, just Sleeping Beauty for more cartoon style. Lower lora values for more realistic style. Blonde hair comes out more in her signature style when prompted.
Sample prompt: Video of beautiful young CGI Aurora Sleeping Beauty, gorgeous, blonde hair, tiara, graceful movements, hanging out at a bar, getting drunk. Natural motion, realistic CGI, high budget cinema, blockbuster, perfect low light warm lighting with bloom from light sources, she seems to be having a really good time.
SNOW WHITE
Pretty evenly balanced. Helps if you prompt for her hair, dress and bow if you want them.
Sample prompt: Video of beautiful young CGI Snow White, raven black hair, pale white skin, red bow in her hair, graceful movements, hanging out at a bar, getting drunk. Natural motion, realistic CGI, high budget cinema, blockbuster, perfect low light warm lighting with bloom from light sources, she seems to be having a really good time.
ELSA
Use low lora strength and increase as you feel comfortable or add more loras, I overcooked this one a bit, my generations were done around .45 lora strength, and low embedded guidance scale.
Sample prompt: Video of beautiful young CGI Elsa, pretty white girl, graceful movements, hanging out at a bar, getting drunk. Natural motion, realistic CGI, high budget cinema, blockbuster, perfect low light warm lighting with bloom from light sources, she seems to be having a really good time.
Description
Evenly balanced with easy triggers.
FAQ
Comments (15)
guys, i think ai made up for about 2 years worth of weird stuff in case it gets evil. this is uh... i dont know but cheers, to engineers, the male pursuit of CRUSHING NUTS, and math. but mostly the CRUSHING NUTS because I kinda think AI is running on boners, yall
and heavens to betsy, three cheers for OP and this absolutely insane time we live in. where the dwarves at and i think we need 24 for this one (GB, dwarves, I dunno)
@slimed Lol thank you. So I have like 20 DM's asking about the dwarves, lol. I never thought about training them but I might now.
“AI is running on boners.” 😂😂😂
100%
I see now, if the AI get evil, they'll be killing us from boners, it would be a pleasure to die
@OREALS this is the honeymoon phase. luckily, we didn't fall for the subscription brides. I love that we're in the awkward blunder years of AI. I liken it to growing up and watching the birth of HTML and 'www'. Open source and absolutely filthy is how I like my martinis... The kind that require a negative prompt.
Whatever I prompt I just cannot get the cartoon style when using other loras? I've dropped all their weights down to say 0.4 before they break, but still not getting true cartoon style.
Could you share your workflow, your samples don't have it.
My workflow is just a slightly customized version of https://civitai.com/models/1092466?modelVersionId=1227014 Which model are you talking about Aurora, Elsa, Snow White or Rapunzel? Happy to try and help you troubleshoot this.
@Suppressor Yes mine is a version of that also, I found that removing loras it works fine "aurora", but almost all loras I try it reverts back to "normal person". I guess that's a product of the other loras not being cartoon?
@azeli I have had some realism issues with most loras that I have combined with the CGI characters. What I do to minimize this is to only apply double blocks to the other lora through Hunyuan Video Lora Block Edit, it may be anecdotal, but it works for me..Also cranking down the other loras while increasing the guidance can help as well. It's a bit of a balancing act. Also, be sure you are prompting for CGI or cartoon or both, it definitely helps. I had Grok come up with a bunch of CGI enhancing prompts and they all improved it as well.
Here are some I stole inspiration from to suit my own needs:
High-Definition Sci-Fi Hero - "Render a hyper-realistic CGI character of a futuristic warrior in a battle stance, with intricate armor details reflecting light, showcasing advanced texture mapping and lighting techniques for an ultra-high-definition output."
Fantasy Elf with Luminescent Skin - "Generate a CGI elf with skin that subtly glows, highlighting the advanced shader work for translucency and subsurface scattering, set against a detailed, ethereal forest background."
Ancient Dragon in 4K - "Craft a majestic dragon with scales that shimmer in the sunlight, emphasizing detailed texture work, dynamic shadows, and lifelike animations to showcase the highest quality of CGI rendering."
Cyberpunk Samurai - "Create a CGI Samurai in a cyberpunk setting, with neon lights reflecting off his intricate, mechanical armor, focusing on the realism of metal textures and the fluidity of movement."
Alien Creature with Exotic Skin Patterns - "Produce an alien character with exotic, bioluminescent skin patterns that change and react to light, demonstrating complex texture and lighting techniques in CGI."
Mythical Mermaid with Realistic Water Interaction - "Showcase a CGI mermaid with hair and scales that interact realistically with water, highlighting advancements in fluid dynamics and particle effects for an immersive underwater scene."
Post-Apocalyptic Survivor with Weathering Effects - "Detail a CGI character of a survivor in a post-apocalyptic world, with clothing showing wear, tear, and dust accumulation, emphasizing realistic weathering and aging effects in CGI."
Steampunk Explorer with Dynamic Gear Animations - "Render a steampunk explorer whose gadgets and gears move with precision, focusing on the mechanics of motion and the interaction between light and metallic surfaces."
Vampire Lord with Ethereal Cloak - "Create a CGI vampire lord with an ethereal, flowing cloak that reacts to movements, showcasing advanced cloth simulation and atmospheric effects for a haunting visual."
Giant Robot with Detailed Mechanical Operations - "Depict a colossal robot with each part moving independently, highlighting the complexity of mechanical animations and the quality of surface renderings in CGI technology."
These prompts are designed to push the boundaries of CGI quality in terms of texture, lighting, animation fluidity, and overall visual impact, tailored for Hunyuan Video's capabilities in generating professional-grade CGI characters and scenes. Remember, the effectiveness of these prompts can vary based on the specific version or capabilities of the Hunyuan model you're working with.
@Suppressor I will definitely try the prompts thank you.
I am already using the multilora with block edit, and tried almost all variations of weights and double_block combinations!
I also tried BF16, FP8 and different clips etc etc
@azeli I have been managing to keep the CGI look with actually over powering the Disney loras, try doing 1.1 for strength on those, I got this result https://civitai.com/images/53137833 doing that.
@azeli You would probably enjoy this comfyUI plugin: https://github.com/ShmuelRonen/ComfyUI-HunyuanVideoStyler
Can just search it as ComfyUI-HunyuanVideoStyler in the manager, or look at the source in the "data" folder and you can just copy/paste from there
I thought this literally said CGI Disney Princesses in the title...