MIMIX

Tingting Liao Chongjian Ge Guangyi Liu Hao Li Yi Zhou
Mohamed bin Zayed University of Artificial Intelligence
arXiv arXiv GitHub Code YouTube Video

🎯 Abstract

✨ Imagine Mr. Bean stepping into Tom & Jerry ✨

Can we generate videos where characters interact naturally across different worlds? We study inter-character interaction in text-to-video generation, where the key challenge is to preserve each character's identity and behaviors while enabling coherent cross-context interaction. This is difficult because characters may have ⚡ never coexisted and because mixing styles often causes 🎨 style delusion, where realistic characters appear cartoonish or vice versa. We introduce a framework that tackles these issues with Cross-Character Embedding (CCE), which learns identity and behavioral logic across multimodal sources, and Cross-Character Augmentation (CCA), which enriches training with synthetic co-existence and mixed-style data. Together, these techniques allow natural interactions between previously uncoexistent characters without losing stylistic fidelity. Experiments on a curated benchmark of cartoons and live-action series with 10 characters show clear improvements in identity preservation, interaction quality, and robustness to style delusion, enabling new forms of generative storytelling.

Mr. Bean appears cartoonish.

Ice Bear appears realistic

Style Delusion Examples

🎬 More Results

Gallery
Grizzly and Panda invite Young Sheldon to their treehouse, where Sheldon tries explaining physics while the bears keep distracting him with snacks.
Mr. Bean is eating spaghetti at a park bench when Panda walks by with a trumpet, playing off-key. The loud noise makes Mr. Bean fling spaghetti in the air. It lands on Panda's head, and they both look at each other in silence.
Ice Bear and Tom join a cooking competition. Ice Bear works quietly with precision, Tom burns everything chasing Jerry across the stove, and George keeps trying to microwave a raw steak. # The background is cartoon style.
Ice Bear calmly paints a picture of Tom, while Tom keeps trying to pose but falls into the paint buckets.
Tom chases Jerry through a city alley but crashes into Ice Bear's shopping cart. Jerry hides in a jar of honey as Ice Bear casually wheels away, ignoring the chaos behind him.
Mr. Bean balances on a rolling office chair while descending a small hill. His tie flaps in the wind as he clutches a briefcase. Panda stands nearby filming the entire stunt on a phone. Mr. Bean waves confidently, then rolls off-screen straight into a bush.
At a quiet library, Mr.Bean sneezes loudly, startling Ice Bear who's balancing a stack of books. The books topple like dominoes, landing with a thud. Both characters exchange a wide-eyed glance before tiptoeing out, pretending nothing happened.
Tom chases Jerry into a library, where Young Sheldon shushes them angrily, but then gets caught in the chaos himself.
Tom hides in Mr. Bean’s closet, but when Jerry joins, they both end up wearing Bean’s oversized clothes.
In a grocery store aisle, Mr. Bean is stacking cereal boxes into a tower. Pada sneezes nearby, causing the tower to collapse. Cereal flies everywhere, and both characters slowly back away, pretending they had nothing to do with it.
Young Sheldon judges a spelling bee, but Panda spells words wrong on purpose, while Jerry sneaks in funny answers.
Young Sheldon visits Mr. Bean’s car, shocked at its strange gadgets, while Jerry sneaks around stealing cookies from the dashboard.
Mr. Bean kneels to pet a robotic dog in a tech store. Tom sneaks up and swaps it with a vacuum cleaner on wheels.
Tom, Grizzly, and Ice Bear form a band on a street corner, with Jerry sabotaging Tom's guitars, sending cheese flying and attracting a crowd of amused pigeons.
Generated Video Sample #15
Young Sheldon Cooper arranges orange traffic cones to run a controlled physics experiment. Grizzley mistakes it for an obstacle course, sprints through at full speed, knocks over every cone, and slides through the grass. Sheldon stares silently, blinking at the chaos before writing down observations.
Mr. Bean blows up a balloon. Jerry hides inside. Mr. Bean pops it and Jerry lands on Mr. Bean's head.
Panda teaches Jerry a dance routine for his livestream, while Tom tries to join but clumsily knocks over the camera and lights.
Young Sheldon Cooper and Panda are sitting on opposite benches in a park, trying to imitate each other's every move. After a series of increasingly ridiculous poses, they both burst out laughing at their synchronized goofiness.

👥 Multi-Character Comparison

✨ Ours
🎬 Skyreel-A2

Prompt:
"Mr. Bean and Tom accidentally get locked inside a fancy hotel kitchen. While Tom chases a mouse under silver trays, Bean tries to cook pasta but ends up flooding the entire room with foam from a dishwasher he mistook for an oven."

Prompt:
"Tom, and Panda go fishing on a rowboat. Tom keeps falling into the water while chasing his bait, Panda takes selfies with each fish, and Tom somehow catches a boot, a sandwich, and a lawn chair—but no fish. "

Prompt:
"Tom and Ice Bear get jobs at a bakery. Tom keeps chasing Jerry through cake trays, and Ice Bear calmly decorates a three-tier wedding cake. The bride ends up choosing Ice Bear’s version over the original."

Prompt:
"Mary Cooper and Panda host a tea party. Mary brings her finest china and proper etiquette. Panda makes cute cupcakes with bear faces."

👤 Single-character Comparison

✨ Ours
🎬 Skyreel
🎨 Wan-I2V
💫 DreamVideo
🎥 VideoBooth

Prompt: "Tom is sitting on the floor with a bowl of milk, drinking it happily while closing his eyes in comfort"

Prompt: "Panda is holding a big milk tea, sipping happily while scrolling through his favorite social media app, completely ignoring the world around him"

Prompt: "Ice Bear is silently cleaning the kitchen with great focus, wearing a small apron, showing his calm and serious personality compared to his brothers"

📚 Citation

BibTeX
@article{mimix2025,
  title   = {Character Mixing for Video Generation},
  author  = {Tingting Liao, Chongjian Ge, Guangyi Liu, Hao Li, Yi Zhou}, 
  year    = {2025}
  eprint       = {2510.05093},
  archivePrefix= {arXiv},
  primaryClass = {cs.CV},
  note         = {arXiv preprint}
}

If you find our work useful in your research, please consider citing our paper.