AnimateDiff works with Initial Images By conditioning the motion module on an image, the motion module can learn to incorporate its composition and context into the final output. Thank you to @mariyaivasileva @advadnoun for collaborating #AnimateDiff

6/ Superstudio is live now. Make it yours.

5/ What’s next? • Upcoming integrations with @Runway and @Kling w/ @FAL • Lightning fast image generation w/ @runware • More customization and fine-tuning • Upscaling for images and videos • New tools for sound and visuals • Real-time collaborative creativity

4/ @LumaLabsAI’s Dream Machine, a powerful text-to-video and image-to-video generator.

3/ Flux is a cutting-edge text-to-image model by Black Forest Labs (@bfl_ml), offering lifelike images and more robust text-in-image capabilities.

2/ Superstudio is a space that brings together powerful generative tools, foundational models, and seamless third-party integrations—starting with Flux and Luma’s Dream Machine.

1/ As a creative on a team of creatives, I knew we deserved more from generative tools. We didn't want to only compete in the image/video gen race anymore. Building a home for creatives mattered more. So we built Superstudio.

Whatever you think we’re working on @KaiberAI is probably wrong. 🙂 Come in and enjoy the show:

squishy keyboard

🙏🏻 @KaiberAI 🤍

more audioreactive visuals All The Words We Don't Say - @OmariJazz Remix Finally got it to evolve instead of returning to the same frame - will likely have both options available. Quite happy with this result! @KaiberAI

Make Anything Audioreactive Using image + prompt + audio, you can make long form visualizers syncing the motion to the audio strength Coming soon to @KaiberAI and will port to ComfyUI when I can finally take a breather

We’re excited to see what you can do with these, and as always you can find these tools available to you at https://t.co/VXtTa5mz8R as well. Stay tuned for more updates.

You can find our repo here: https://t.co/jaIPT5Z68f If you'd like to discuss research, interact with the team, propose experiments, please come visit our Discord!

This repo currently has a modified version of init image, which allows you to create stylized versions of the input. The examples above will require a slightly modified architecture and trained checkpoint, which will we release soon. An example of the current version is below:

Our first contribution is the publicization of our enhanced AnimateDiff repository. This repository comes with a range of enhancements, including initial image support and high res fixes. These upgrades aim to enhance the possibilities of AnimateDiff.

Here at Kaiber, we are always grateful for the open source AI community, which has provided us with invaluable resources and support throughout our journey. That is why we are excited to contribute back to the community, with the start of our own set of open source initiatives.

You can string this together end to end to further lengthen the videos, creating a fully autoregressive video generation system.

We are hiring a generative AI expert to lead and build our creative research team @KaiberAI 🚀 🌎 remote but we party 🎨 deep passion for the arts ✨work with superstar team and clients https://t.co/ndmqjBAi79