Takes a realistic or 3D frame as the initial image, then converts it to an animated style. This LoRA by itself may yield a wide variety of anime-styled to semi-realistic cartoon appearances, but it can be directed to more reliably produce anime by pairing it with my Anime Style LoRA.
Example Use:
Strength: 1.5 (on both this LoRA and the Anime Style LoRA)
Length: 33 frames
Prompt: "a realistic person shuffles in place, later becoming An1meStyl3"
Steps: 4 (2 high, 2 low)
California AB 2013 Training Data Disclosure
This LoRA was fine-tuned using visual data consisting of still images derived from animated works, along with photographs, game screenshots, and AI-generated images used as inputs for synthetic transformation. The training data includes copyrighted material owned by third parties, including animation studios, game publishers, and individual artists, as well as derivative synthetic content generated from those inputs. No training data was licensed or purchased. This LoRA is provided for non-commercial use only under the terms of its distribution.
The dataset consists of over 100 images derived from anime-style sources, along with over 50 video clips. Each clip was created by taking an existing image, including photographs, game screenshots, or AI-generated images, generating an anime-style version of that image using AI, and producing an animated sequence in which both versions appear sequentially to introduce a deliberate style transition. Data was collected and generated between approximately 2024 and 2025.
Image data was processed through standard resizing, cropping, normalization, and labeling steps. Video clips were prepared and labeled for video-based training to support motion learning and style-transition behavior. Synthetic images and video were included as part of the training dataset.
This model is intended for non-commercial, experimental, and educational use. Generated outputs may reflect copyrighted visual styles or themes associated with the underlying training data. Users are responsible for ensuring compliance with applicable copyright law, other intellectual property laws, and all other applicable laws.
Description
FAQ
Comments (11)
Hi, im currently also working on "blink apply style" lora.
Just wondering which dataset did you make?
1 frame in realism and others anime?
Hi, what will be the concept of your blink lora?
The version I posted is actually my third training attempt; on the earlier versions, I tried to have the clips be 18 frames with 18 frames of context, switching exactly at frame 10, and it wasn't effective. On this version, I used two datasets in Ostris:
1) 60 video samples that showed the transition, with 10-15 frames of each (ensuring the window changes but the change always fits within my training window), but then trained with a 20 frame context window [dataset weight: 1.0]
2) 300 image samples that showed the desired output style (fully tagged using "An1meStyl3") [dataset weight: 0.1]
I just posted a few of the real data samples for reference: https://civitai.com/posts/25328233
@Rating_Agent 2.5D digital comics style
I want to add my favorite After Effects Post-Production Spice
It will be fun and interesting
@Polymath_wtf I ask you to please archive your Lora’s on huggingface too, civil admins hate wan Loras
Thank you, Great contribution.
Is it possible to Realistic image to Direct anime style.
I mean, instead of a realistic image switch to anime, turn to an anime style directly.
Potentially; I fear that I may not understand the question. Do you mean like generate a video directly as an anime without a realistic reference image? If so, this LoRA cannot do it, but my T2V anime one can: https://civitai.com/models/2200950/anime-style-wan-22-t2v
@ReltivlyObjectv i had a similar question. i would like to know if it could be possible to skip the transition, so the first generated image is already anime.
@FortunaCournot Not that I am aware of. WAN I2V models operate temporally, so if you're wanting a similar frame you could experiment with prompts along the lines of "the person holds still." If you need the exact same frame but anime, I believe something like Qwen Image Edit would be the best path forward.
This is an absolutely fascinating LoRA! Having a lot of fun playing with it and observing how it interacts with Wan. Great stuff.
Thank you very much for this 🫶
Need to add as a feedback report, that to see things really changing on my end I needed to volume up high and low loras on this one up to 2.0-2.5 for "Switch to anime" loras and 1.5-2.0 for the "anime style" loras. (working with the GGUF Checkpoint "Synth Seduction" from "Darksidewalker")
Also I found the transition steps very lovely, so I'm trying to get a solid terminus for this gem of a halfway-mix-style right now. Luv it ! : ))