Junshu Tang

Senior Researcher in Hunyuan / PhD. in SJTU

[Junshu Tang @tangjs9707]

No more catastrophic forgetting. Just smarter, instance-level adaptation!! HY-WU uses Functional Neural Memory to scale intelligence. Congrats on @VictorKaiWang1!!🥳

[Tencent Hy @TencentHunyuan]

One static model does not fit all😭

We just dropped our latest work: Functional Neural Memory. Instead of static models, we generate custom "parameters" for every single input.

✅Prompt your model anytime
✅Instant personalization
✅Better instruction following
✅Flexible & dynamic memory (w/o memory bank✌️)

(🧵1/6)

[Tencent Hy @TencentHunyuan]

One static model does not fit all😭

We just dropped our latest work: Functional Neural Memory. Instead of static models, we generate custom "parameters" for every single input.

✅Prompt your model anytime
✅Instant personalization
✅Better instruction following
✅Flexible & dynamic memory (w/o memory bank✌️)

(🧵1/6)

[clem 🤗 @ClementDelangue]

Cool to see an open world model number 5 trending models on HF and a 250M parameter model number 3.

Starting to be in a post-LLM world!

[Tencent Hy @TencentHunyuan]

🚀We are thrilled to open-source Hunyuan-GameCraft, a high-dynamic interactive game video generation framework built on HunyuanVideo.

It generates playable and physically realistic videos from a single scene image and user action signals, empowering creators and developers to "direct" games with first-person or third-person perspectives.

Key Advantages:
🔹High Dynamics: Unifies standard keyboard inputs into a shared continuous action space, enabling high-precision control over velocity and angle. This allows for the exploration of complex trajectories, overcoming the stiff, limited motion of traditional models. It can also generate dynamic environmental content like moving clouds, rain, snow, and water flow.
🔹Long-term Consistency: Uses hybrid history condition to preserve the original scene information after significant movement.
🔹Significant Cost Reduction: No need for expensive modeling/rendering. PCM distillation compresses inference steps, boosting speed and lowering costs. This allows the quantized 13B model to run on consumer-grade GPUs like the RTX 4090.

Project Page: https://t.co/uAbiu9FRzF
Code: https://t.co/WgppVz1KUq
Technical Report: https://t.co/aO8plomaTr
Hugging Face：https://t.co/2ZOUWm6KKQ

[Junshu Tang @tangjs9707]

Thanks @_akhaliq for tweeting our work!🤩🤩 GameCraft explores spatial intelligence in video models, which can maintain the dynamics of characters and environments while also having good long-term and 3D consistency. 🏡

[AK @_akhaliq]

Hunyuan-GameCraft

High-dynamic Interactive Game Video Generation with Hybrid History Condition

[Junshu Tang @tangjs9707]

#ECCV2024 Our TexDreamer is accepted as Oral !!! 🥳

[Junshu Tang @tangjs9707]

Glad to announce that our TexDreamer is accepted to #ECCV2024 !!🥳🥳 And welcome to follow our newest work ClotheDreamer which works on text-guided garment generation!!💃🕺
TexDreamer:https://t.co/2NAw47w4Zr
ClotheDreamer: https://t.co/7gWHP4KcJt

[Junshu Tang @tangjs9707]

Glad to announce that our TexDreamer is accepted to #ECCV2024 !!🥳🥳 And welcome to follow our newest work ClotheDreamer which works on text-guided garment generation!!💃🕺
TexDreamer:https://t.co/2NAw47w4Zr
ClotheDreamer: https://t.co/7gWHP4KcJt

[AK @_akhaliq]

TexDreamer

Towards Zero-Shot High-Fidelity 3D Human Texture Generation

Texturing 3D humans with semantic UV maps remains a challenge due to the difficulty of acquiring reasonably unfolded UV. Despite recent text-to-3D advancements in supervising multi-view renderings

[Junshu Tang @tangjs9707]

Another amazing multi-person motion generation work:"👯‍♂️FreeMotion: A Unified Framework for Number-free Text-to-Motion Synthesis" is also accepted to #ECCV2024!!🤩🤩
Project page: https://t.co/JAjjXxd5gf

[MrNeRF @janusch_patas]

ClotheDreamer: Text-Guided Garment Generation with 3D Gaussians
https://t.co/i5lWz74JyT
Project: https://t.co/CKS8boywDI

[AK @_akhaliq]

TexDreamer

Towards Zero-Shot High-Fidelity 3D Human Texture Generation

Texturing 3D humans with semantic UV maps remains a challenge due to the difficulty of acquiring reasonably unfolded UV. Despite recent text-to-3D advancements in supervising multi-view renderings

[Junshu Tang @tangjs9707]

Thanks @_akhaliq for tweeting our work!🥰
Glad to announce that our Make-It-3D has officially been accepted to #ICCV2023🥳🥳🥳
Congrats to all the best co-authors @DylanTFWang @zhangboknight ~
Make-It-3D is now open-source on GitHub https://t.co/DcHbBP2V00 and get 1k+ 🌟!!

[AK @_akhaliq]

Make-It-3D: High-Fidelity 3D Creation from A Single Image with Diffusion Prior

abs: https://t.co/qykLHelrjP
project page: https://t.co/ZumkSC4V0N

[AK @_akhaliq]

Make-It-3D: High-Fidelity 3D Creation from A Single Image with Diffusion Prior

abs: https://t.co/qykLHelrjP
project page: https://t.co/ZumkSC4V0N

[AK @_akhaliq]

Explicitly Controllable 3D-Aware Portrait Generation
abs: https://t.co/jIq7V7Ijpd
project page: https://t.co/Mnms3MgxS6