yongzhang6782

> > I followed the paper for re-implementation. I got good results on DF and FS, but pool results on F2F and NT. Did you encounter a similar situation? >...

> > > > I followed the paper for re-implementation. I got good results on DF and FS, but pool results on F2F and NT. Did you encounter a similar...

> > > > I followed the paper for re-implementation. I got good results on DF and FS, but pool results on F2F and NT. Did you encounter a similar...

> Hi, @yzhang2016 @jerry4h . My implementation also heavily overfits BI dataset. I also notice that in my experiments the noise difference in fake images from BI dataset is much...

我也遇到通用的问题。 imdb 和 wiki 有大量的错误标注，而且很多年龄并不准确。这可能是导致训练效果较差的原因之一。另外，年龄的不均衡，导致预测基本在训练集中占比较大的年龄段。对小孩的预测非常不准。

Our new I2V model with the resolution 640x1024 has been released. Much better then the previous version. Feel free to try it.

News: we release a new version of VideoCrafter on Discord/Floor33 (https://discord.gg/uHaQuThT). It has a significant improvement over the previous one. It can generate 720p film-like videos. Please join us on...

> Do you have any good solutions for the flickering issue in generated videos? We release a new version on Discord/Floor33 (https://discord.gg/uHaQuThT). It works much better in terms of flickering.

Yes. OMG is better. On Wed, Aug 28, 2024 at 12:52 AM John D. Pope ***@***.***> wrote: > is the follow up work better? https://github.com/kongzhecn/OMG > > — > Reply...

InstantID can take multiple images as reference. The embeddings of all reference images are averaged as the input of the identity net (control net).