Anime2Reality ComfyUI Workflow
This workflow transforms anime-style images into high-resolution, realistic human photos. It uses Qwen Edit as the base model for detailed anime-to-human conversion and WAN as a high-resolution refiner to enhance the output.
Key features:
High-resolution outputs: Achieve crisp, photorealistic results with minimal artifacts.
Custom latent upscaler: Includes a proprietary upscaler node to boost quality beyond standard upscaling techniques. GitHub link
Stepwise refinement: Combines base conversion and high-res refinement in multiple stages for consistent, lifelike results.
Turn your favorite anime characters into stunningly realistic human portraits with precise control over resolution, pose, and detail.
Description
FAQ
Comments (10)
Failed to validate prompt for output 152:
* LoraLoaderModelOnly 143:111:
- Value not in list: lora_name: 'wan\Wan2.1_T2V_14B_FusionX_LoRA.safetensors' not in (list of length 547)
I cannot find this node in the workflow.
Edit: found it I had to unpack the wan loader node.
iPiKo here is my latent scaler:
https://github.com/DenRakEiw/WAN_NN_Latent_Upscale
denrakeiw the problem was not with the node there, you do not have an option to reselect the wan fusion lora outside the new feature what is it called graph? So peoples have to figure it out.
iPiKo Yes, maybe it’s not such a good idea to build the loader nodes as a subgraph.
denrakeiw it's not bad idea I like it this way but you just forgot to make an input for the fusion lora for user to adjust it.
Small issue on the github page, the installation says to "git clone https://github.com/yourusername/wan_nn_latent.git"
whereas "https://github.com/DenRakEiw/WAN_NN_Latent_Upscale.git" would be the correct path.
getting error: "Prompt outputs failed validation:
FluxEmptyLatentSizePicker:
- Value not in list: initialization_mode: '0' not in ['zeros', 'vae_sample', 'gaussian_noise', 'uniform_noise']
- Failed to convert an input value to a INT value: batch_size, 3:4, invalid literal for int() with base 10: '3:4'
- Failed to convert an input value to a FLOAT value: noise_strength, zeros, could not convert string to float: 'zeros'
- Value not in list: resolution: '1.5MP' not in (list of length 24)"
Circular reference detected while resolving output 0 of node 146 (CLIP Text Encode (Positive Prompt)) at path 146. This creates an infinite loop in link resolution.
I get that MSN error help
I ran into some issues. The FusionX Lora is locked inside the Wan Loader SubGraph, meaning the user will get errors if it isn't that exact name and path, and won't be able to find the place to fix it.
After fixing that, I am stuck on this error:
Prompt outputs failed validation: ModelSamplingAuraFlow: - Failed to convert an input value to a FLOAT value: shift, simple, could not convert string to float: 'simple' UniversalNNLatentUpscale: - Failed to convert an input value to a FLOAT value: upscale, Flux, could not convert string to float: 'Flux' UniversalNNLatentUpscale: - Failed to convert an input value to a FLOAT value: upscale, Flux, could not convert string to float: 'Flux'