deadman44

.title{ font-size: 2.5em; letter-spacing: 0.01em; padding: 0.5em 0; } .thumbwidth{ max-width: 180px; } .fontred{ color:red; } .fontblue{ color:blue; } .fontgrey{ color: #aaaaaa; } models - ZipatriousXL1536testd ( Illustrious Base ):2025-06-13 - ZipatriousXLv1.0b ( Illustrious Base ):2025-04-19 - ZipanoobXLEpredv1.1 ( Noob AI Base ):2025-01-31 - ZipanoobXLVpredv1.1 ( Noob AI Base ):2025-01-14 - Zipang XL test03.1 ( Animagine Base ): 3.1b:2024-07-22 - PonypangXLgiveup ( Pony Base ): fix5:2024-07-03 - Experimental Version - 1536x1536 Finetune trained - The number of datasets has been reduced because we trained only with high-resolution images. Download: 1536testd -Continued learning based on testc, but not good. (fail) Download: 1536testc -another approach. yo tag might be soft on Download: 1536testb -20000+ images Finetune trained - Experimental Version - More realistic - Expression is not wide - Many bankruptcies - Animated characters cannot be output. - dmd2sdxl4steplora included - ADetailer or HiresFix Recommendation. Download: v1.0b (only increased the number of learning steps) Euler a SGMUniform 14-30 steps CFG Scale:1.3 Download: v1.0 Euler a Uniform 10 steps CFG Scale:1.3 # - base model - Illustrious-xl-early-release-v0 -great XL model! -(Epred) Base model:NoobAI XL Epsilon-pred 1.1-Version -(Vpred) Base model:NoobAI XL V-Pred 1.0 - dmd2sdxl4steplora included - ADetailer or HiresFix Recommendation. (Vpred version) -(forge) turn on Zero Terminal SNR Turn off when using flux or other epred models # - base model - NoobAI XL Epsilon-pred 1.1-Version - NoobAI XL V-Pred 1.0 -great XL model! -4000+ twitter images trained & 10000+ images merged model -Animagine XL 3.1 Base -Good Lighting -photoreal like tag: shadow, flash lighting, backlighting, silhouette, sunset, night, day, bokeh, etc. Download:test3.1 (Recommended) Download:test3.1b (Newer isn't always better) Sample prompt -SDXL Lightning (too fast: recommended) Mixed LoRA due to problems with training not being able to continue. e621 tag might be better than danbooru tag. -4000+ twitter images trained & 10000+ images merged model -experimental -Might look like Zipang. -Hand is not good. -ADetailer and HiresFix Recommendation. - base model - Pony Diffusion V6 XL -great XL model! - ebara pony 2.1 -great XL model! too

.title{ font-size: 2.5em; letter-spacing: 0.01em; padding: 0.5em 0; } .thumbwidth{ max-width: 180px; } .fontred{ color:red; } .fontblue{ color:blue; } .fontgrey{ color: #aaaaaa; } models - Wan2.2myjsv02 ( Wan2.2 LoRA ):2025-09-17 - Wan2.2myjcv02 ( Wan2.2 LoRA ):2025-09-04 - Wan2.2myobv01 ( Wan2.2 LoRA ):2025-08-31 - Wan2.2myjdv01 ( Wan2.2 LoRA ):2025-08-26 - Wan2.2myjyv01 ( Wan2.2 LoRA ):2025-08-21 - Wan2.2myjkv01 ( Wan2.2 LoRA ):2025-08-18 - Add Workflow page: 2025-08-04 - Workflow ### - Sample Workflow for myxx series LoRA - Wan video 2.2 for t2i, t2v for ComfyUI - The optimal resolution is 768 x 1280, 1024 x 1536.(T2i), 512x768 (T2v) - natural Japanese OB face Download Download: myobHighv01 Download: myobLowv01 - Wan video 2.2 for t2i, t2v for ComfyUI - The optimal resolution is 768 x 1280, 1024 x 1536.(T2i), 512x768 (T2v) - natural Japanese JD face Download Download: myjdHighv01 Download: myjdLowv01 - Wan video 2.2 for t2i, t2v for ComfyUI - The optimal resolution is 768 x 1280, 1024 x 1536.(T2i), 512x768 (T2v) - natural Japanese JK face Download Download: myjkHighv01 Download: myjkLowv01 - Wan video 2.2 for t2i, t2v for ComfyUI - The optimal resolution is 768 x 1280, 1024 x 1536.(T2i), 512x768 (T2v) - natural Japanese JC face Download Download: myjcHighv02 Download: myjcLowv02 - Wan video 2.2 for t2i, t2v for ComfyUI - The optimal resolution is 768 x 1280, 1024 x 1536.(T2i), 512x768 (T2v) - natural Japanese JS face Download Download: myjsHighv02 Download: myjsLowv02 - Wan video 2.2 for t2i, t2v for ComfyUI - The optimal resolution is 768 x 1280, 1024 x 1536.(T2i), 512x768 (T2v) - natural Japanese JY face Download Download: myjyHighv01 Download: myjyLowv01

.title{ font-size: 2.5em; letter-spacing: 0.01em; padding: 0.5em 0; } .thumbwidth{ max-width: 180px; } .fontred{ color:red; } .fontblue{ color:blue; } .fontgrey{ color: #aaaaaa; } Recent - I2v better workflow high has been replaced with smoothmix (it works better, but if the video is unstable, please use i2v triple) - Replaced the block swap in the Triple High node with multi-GPU Node. (Reportedly speeds things up by about 10%) - Fix I2v Triple High Workflow (wrong node): 2025-08-29 - Add T2v Triple High Workflow / I2v Triple High (experimental): 2025-08-21 - flf2v has been integrated into I2vbetter: 2025-08-16 - Added Video Extend function to I2vbetter Workflow: 2025-08-15 - Implement "my series LoRA select switch" on each node: 2025-08-14 - Add Interpolation Workflow / Update some nodes: 2025-08-13 - T2i / T2v Workflow - I2v Workflow - Highres Upscale Workflow - Speed Mult Workflow - Concat Workflow - Interpolation Workflow - Reccomended Models T2v Triple High - better motion quality, but very slowly 15yo, myjc, japanese, The camera approaches a girl in school uniform walking at street of Tokyo. She has a black twintails with side bangs. The girl smiles and waves at the camera, then turns around and runs away. The camera moves away from the girl. 15yo, myjc, japanese, photorealistic, portrait of a girl walking at street. She has a long black ponytail with side bangs. She is wearing short sleeves school uniform. T2i / T2v - Better Quality Workflow - Recommended for T2i 15yo, myjc, japanese, photorealistic, An upper body portrait of a girl in white dress at library. She has a long black hair with bangs. She is holding books, and looking at camera. A girl taking a selfie. The camera circles around the girl. She has black hair, a ponytail, and is wearing a school uniform. The background is a daytime classroom with several students. I2v - Better Quality Workflow - Recommended for I2v / flf2v / Extend - I2v High: smoothmix I2v low: Q8 gguf Extend: fun 18yo, myjk, japanese, A nurse is walking down a dark hospital corridor when suddenly the doctors appear with a bouquet of flowers and she jumps up and down, smiling and overjoyed. Cut Start Frames - Reduces discomfort at edges, etc. (e.g., collage images) by removing the initial few frames. Video Extend - Overlapping video frames ensures consistent extension - Load Image a "webp video file". - Wan2.2-Fun-A14B-InP-GGUF required Last Image + I2V - T2V 5sec (81frames) + LastFrame I2V 5sec (81frames) - The speed changes when switching videos. Video Extend - (Base) T2V 5sec (81frames): 12frames overrap 10sec - The transition is smooth. 18yo, myjk, japanese, Suddenly A girl in a yellow parachute suit lands with her parachute open on a street crowded with people. She has a black ponytail with side bangs. People cheer and applaud. 18yo, myjk, japanese, A girl is running. She wears black twin-tail hair and a school uniform. In the background is a row of trees and a blue sky, her hair blowing in the wind. I2v / flf2v / Extend Triple High - better motion quality, but very slowly I2v / StartEnd (kijai node) - Recommended for Last Image + I2v 15yo, myjc, japanese, photorealistic, Suddenly few students appear. The girl is dancing with students. Highres Upscale Workflow - High-resolution upscaling of load images and videos - image file and movie (webp) - Low denoise shows less change, high denoise shows more change Speed Mult Workflow - Change video speed - webp movie -> mp4 # Concat Workflow - Connecting videos - (webp) Videos with different fps don't work well together. image -> Highres (webp) -> SpeedMult (mp4) -> Concat (mp4) # Interpolation Workflow - Frame Interpolation videos - webp / mp4 -> fps:60 image -> Highres (webp) -> SpeedMult (mp4) -> Concat (mp4) - Wan2.2 Q5, Q4, Q3... will further omit vram [T2i / T2v] - wan2.2t2vhighnoise14BQ80.gguf - wan2.2t2vlownoise14BQ80.gguf [I2v] - smoothMixWan22I2VT2Vi2vHigh.safetensors - wan2.2i2vhighnoise14BQ8KM.gguf - wan2.2i2vlownoise14BQ8KM.gguf [I2v Video Extend] - Wan2.2-Fun-A14B-InPHighNoise-Q80.gguf - Wan2.2-Fun-A14B-InPLowNoise-Q80.gguf [etc] - umt5xxlfp8e4m3fnscaled - umt5-xxl-enc-bf16.safetensors - Wan21VAEbf16.safetensors [LoRA] - wan2.2i2vA14bhigh/lownoiselorarank64lightx2v4step1022

.title{ font-size: 2.5em; letter-spacing: 0.01em; padding: 0.5em 0; } .thumbwidth{ max-width: 180px; } .fontred{ color:red; } .fontblue{ color:blue; } .fontgrey{ color: #aaaaaa; } models - Add loraqwenmyjcv03 ( Edit-2509 LoRA ):2025-11-25 - Add loraqwenmyjsv02b ( Edit-2509 LoRA ):2025-10-28 - Add loraqwenmyjyv01 ( Edit-2509 LoRA ):2025-10-19 - Add loraqwenmyjkv02 fix ( Edit-2509 LoRA ):2025-10-17 Sample Workflow - Workflow for myxx series LoRA - reccomended - Workflow for Double - Pretty good but slow - Workflow for Triple - Mostly bad, but sometimes good - Workflow Image Edit - Qwen-Image-Edit-2509 workflow - reccomended models - Qwen-Image-gguf - Qwen-Image-Edit-2509-GGUF - Text Encoder - VAE - Lightx2v LoRA --- - Lora for Qwen-Image / Edit-2509 - natural Japanese JC face Download Download: myjcv03 (Edit-2509) Download: myjcv02 (Edit-2509) Download: myjcv01 (Qwen Image) Download Download: myjsv02b strength:2.5 might be better Download: myjsv02 Download: myjsv01