I thought I would add this here since there's 100+ issues asking for code. FYI, I'm not affiliated with this paper in any way.
First of all, nowhere in the paper do the authors actually promise code, so they may not ever do so. I hope they do, it's definitely possible to reimplement this work, but they use a lot of data and compute power that would be hard for smaller groups to copy. There's a good chance they will as they've made the repo and have suggested in the past that the similar work, animate anyone, will eventually be released.
If they do choose to release it, the authors have almost certainly submitted this paper to ECCV (https://eccv2024.ecva.net), you can tell by the awful single-column format used in the arxiv version. Usually, you can expect code to come out if, and only if, the paper is accepted. Even then it's usually around the camera-ready deadline, this is July 15th in this case (https://eccv2024.ecva.net/Conferences/2024/Dates).
The paper is good but there's no guarantee of acceptance. It may have to go through multiple conferences, meaning the code release would be even later.
The authors can't tell you this because they are expressly forbidden to mention their paper is an ECCV submission by the author guidelines. This is supposed to be to maintain anonymity (which is, of course, impossible with a paper gaining this much hype but that's another issue).
fengxiang777
commented
4 months ago
I thought I would add this here since there's 100+ issues asking for code. FYI, I'm not affiliated with this paper in any way.
First of all, nowhere in the paper do the authors actually promise code, so they may not ever do so. I hope they do, it's definitely possible to reimplement this work, but they use a lot of data and compute power that would be hard for smaller groups to copy. There's a good chance they will as they've made the repo and have suggested in the past that the similar work, animate anyone, will eventually be released.
If they do choose to release it, the authors have almost certainly submitted this paper to ECCV (https://eccv2024.ecva.net), you can tell by the awful single-column format used in the arxiv version. Usually, you can expect code to come out if, and only if, the paper is accepted. Even then it's usually around the camera-ready deadline, this is July 15th in this case (https://eccv2024.ecva.net/Conferences/2024/Dates).
The paper is good but there's no guarantee of acceptance. It may have to go through multiple conferences, meaning the code release would be even later.
The authors can't tell you this because they are expressly forbidden to mention their paper is an ECCV submission by the author guidelines. This is supposed to be to maintain anonymity (which is, of course, impossible with a paper gaining this much hype but that's another issue).