Moves from Adobe, Microsoft/GitHub, Unity, and Text-to-Video Innovations.
The generative AI landscape is evolving at breakneck speed, with groundbreaking advancements announced almost daily. 🌊
In this week’s roundup, we’ll explore two key themes: major moves from Adobe, Microsoft/GitHub, and Unity, as well as how text-to-video technology is becoming more powerful and consistent.
Adobe Firefly suite of generative models 🚀
Though Adobe has offered generative AI functionality like face editing for some time, they’ve recently made a significant leap into the world of transformer-based models introducing Firefly, a suite that includes text-to-image, text masking, and more features in the pipeline. The text masking tool is particularly noteworthy, resembling the Stable Diffusion and controlnet capabilities used to generate the BIFF logo. Discover more about Firefly here.
GitHub’s ChatGPT-Inspired Copilot X 🤖
Not long ago, I wrote how GitHub’s Copilot felt outpaced by ChatGPT. It seems many shared this sentiment, as GitHub recently announced their ChatGPT-like GitHub Copilot X. I’m eager to sign up, fire up NeoVIM again and try it out when it is available, but for now we can learn more about it here.
Unity Embraces Natural Language AI in Gaming 🎮
Unity, the leading game development platform, has integrated natural language AI, hinting at the transparent interfaces of the future. Check it out here.
The Surge of Text-to-Video Innovations 🎥
The text-to-video branch of generative AI is experiencing a surge of incredible innovation. With reduced flickering and enhanced quality, this technology will soon empower creators to conjure fantastic visuals with minimal effort and cost.
This week’s standout news includes the release of Modelscope, a research text-to-video model from Alibaba’s team, offering weights and examples. While outputs are strange and unpredictable, this model is a major step forward in consistency for open models. Explore Modelscope here.
Meanwhile for closed models, Runway Gen-2 text-to-video is looking impressively sharp! Take a look at it here.
Finally, an outside contender: Instruct-NeRF2NeRF, a Neural Radiance Fields implementation of text generation. By integrating Instruct Pix2Pix into the NeRF training process, this technology enables the creation of natural language-editable 3D scenes and effectively rendered video. Learn more here.
That’s it for this week’s update. Stay tuned for more exciting developments in the coming weeks!