stable-diffusion-webui-rembg VS ECCV2022-RIFE

Extension REMBG for Automatic1111

This tool will mask the output image of your generations: https://github.com/AUTOMATIC1111/stable-diffusion-webui-rembg

Remove background with your favourite image editing software (I used Photoshop) or rembg, for example

With the following prompt, we get a good image of Pikachu in the DD style which can then have the background made transparent with https://github.com/AUTOMATIC1111/stable-diffusion-webui-rembg

Also, this auto removes backgrounds

So, the only difference from the listed method is to add an extra step to preprocessing. Before captioning images, I used this extension to batch remove backgrounds. Then, I took those PNGs and used Photoshop to batch save them as JPEGs, resulting in cutout images of the subject with white backgrounds. I then proceeded as listed in that comment.

Check out RIFE.

Using Flowframes with the RIFE model, run 2x interpolation on a folder of video frames.

Personally I have mine set to use RIFE AI via TensorRT for frame interpolation(x2), if the FPS is 30 or less.

Oh now that I saw this comment, I started to investigate frame interpolation techniques using AI and found this: https://github.com/megvii-research/ECCV2022-RIFE

Yes but you need to use a different AI first. There are multiple AI like RIFE (there are apps for it if you don't like code) that will smooth out your animation. Then you can use those frames with Novel AI to get a more organic look in the end.

To have more smooth effect you can reduce frame per second and add FILM or RIFE between frames.

Interpolated with https://github.com/megvii-research/ECCV2022-RIFE

Sure! This would be my approach (and tools) if I was smarter:

If you make the generations with some similarities and use the right interpolation, you don't need 1000 images like my video and can obtain a smooth movement.

First, generate images with some kind of visual anchor (background, an object). You can use frames generated using the previous frame as reference image, or the same seed but different prompt/parameters, or you can go wild using img2img/inpainting (btw I struggle to find an inpainting tool for Stable Diffusion: they seem to be just img2img with a mask, without contest).

Then pass the generated images to one of the most recent interpolation algorithms, like this one https://github.com/megvii-research/ECCV2022-RIFE or the one used in the replicate we are commenting on (someone posted this reference: https://github.com/google-research/frame-interpolation )

The first link reports some free and paid implementation and a Colab, so depending on how deep you want to go, you have a lot of choices.

In the end, I'd use some good app to stabilize the image if needed, to get a more "calm" look. I use Luma Fusion, but it's a paid app (cheap, one-time payment, for iOS). I'm sure there are a ton of open-source implementations.

It's an approach similar to the animation on replicate, but it allows a lot of fine-tuning and you can add new animation ideas/tools to the process.

Nothing revolutionary, but I hope it helps!

> You have generated some pretty cool designs.

Thanks! I put in a lot of work in the last weeks. The project has a mission, I wrote something, but it's not ready yet. I believe it will be with the launch of Dall-E 8 :-/

You'd probably get way better results by using something like RIFE to do interpolation and recreate missing frames, instead of minterpolate. I understand though that it's more effort as you'll need to install and setup RIFE.