search

Fictionarry / TalkingGaussian

[ECCV'24] TalkingGaussian: Structure-Persistent 3D Talking Head Synthesis via Gaussian Splatting
https://fictionarry.github.io/TalkingGaussian/
213 stars 28 forks source link

Is this better than synctalk? #44

Open linhcentrio opened 3 days ago

linhcentrio commented 3 days ago

Is this better than synctalk?

Fictionarry commented 2 days ago

Both ours and SyncTalk are good works, while these two aims to solve different problems. SyncTalk solves some "devil" problems and make validations on ER-NeRF. Our work focus on the modeling methodology itself and relatively less on the details, while in the implementation we also use some good design in SyncTalk to provide better performance, on the premise of fairness.

linhcentrio commented 2 days ago

Both ours and SyncTalk are good works, while these two aims to solve different problems. SyncTalk solves some "devil" problems and make validations on ER-NeRF. Our work focus on the modeling methodology itself and relatively less on the details, while in the implementation we also use some good design in SyncTalk to provide better performance, on the premise of fairness.

thank you, it's good new. good job!