TVCG 2026: MARRS for Human Motion Action-Reaction Synthesis

# MARRS: Masked Autoregressive Unit-based Reaction Synthesis

Project page: **https://aigc-explorer.github.io/MARRS/**

Introducing MARRS: a new framework for human action-reaction synthesis that generates coordinated, fine-grained reactions conditioned on another person’s motion. By avoiding VQ and modeling body/hand units with UD-VAE + ACF + MUM, MARRS captures cross-unit perception more effectively and efficiently. It achieves state-of-the-art quantitative and qualitative results.

Overall framework

Demo

https://redd.it/1t853hg
@rStableDiffusion

Has anyone tried LTX2.3 for Image Gen?

Before I moved to ZIT, I used Wan for generating images and it worked quite well. Im wondering if anyone has tried with LTX and if the results were good.

https://redd.it/1t81hrg
@rStableDiffusion

Trained a Vit model from scratch for auto tagging

I recently trained a new anime image tagging model. To prep the data, I used SmilingWolf v3 to fix 300k bad tags and fill in 1M missing ones. I also trained an initial baseline model to help identify and add around 30k low-frequency tags.

The current V1 model is a 320x320 ViT. V1.1 is currently training at 448x448, and the higher resolution is already improving accuracy. My next goal is to wait for a 2025 dataset, clean it heavily, and train from scratch with better vocab structures (e.g., artist:name).

You can find the model, card, and demo space on HuggingFace: https://huggingface.co/Grio43/OppaiOracle Live use of the model: https://huggingface.co/spaces/Grio43/OppaiOracle

CPU based tagger
https://huggingface.co/spaces/Grio43/OppaiCPU

https://redd.it/1t8bzb3
@rStableDiffusion

Its still nuts to me how realistic AI is getting, incredible i can run it on a RTX2060 and get these results. (Z-image-Turbo)

https://redd.it/1t8ehyj
@rStableDiffusion

Teal Dark - Flux.2 Klein 9b style/aesthetic LORA

https://redd.it/1t8bn7f
@rStableDiffusion

Anyone else using LTX locally on Mac via Draw Things? Here’s a WWII-style short I made.

https://redd.it/1t8lagy
@rStableDiffusion

How I feel after upvoting a post that got downvoted by bots for mentioning Forge Neo.
https://redd.it/1t8oha2
@rStableDiffusion

Wan 2.2 with LTX 2.3 ID-LoRA

Wan 2.2 with LTX 2.3 ID-LoRA workflow

This is a workflow that combines the Comfy Wan 2.2 image-to-video workflow with the Comfy LTX 2.3 ID-LoRA workflow. You can use Wan 2.2 to make your initial video then it will automatically run through LTX 2.3 to add audio to your Wan 2.2 video and extend the Wan 2.2 video with whatever you want to happen next.

Wan 2.2 image-to-video of Crystal Sparkle throwing a champagne bottle against a yacht to christen the yacht

LTX 2.3 adds the foley audio to the Wan 2.2 clip for bottle smashing against boat and ID-LoRA adds Crystal Sparkle's actual voice

Here is a link to the workflow: https://huggingface.co/ussaaron/workflows/blob/main/wan2\_2\_i2v-with-ltx-id-lora.json

https://redd.it/1t8qloh
@rStableDiffusion

Hi-Dream 01 Out : 2k Images in 20seconds on a 4090 (fp8 dev) ComfyUI

https://redd.it/1t8ypmd
@rStableDiffusion