AI Art Weekly #127

Hello, my fellow dreamers, and welcome to issue #127 of AI Art Weekly! 👋

This has been one of the best weeks for Midjourney users in quite a while. MJ released their first video model and a new SREF system, so I can keep feeding my generative AI addiction as if I didn’t have work to do. Oh well 😅

Having said that, I’ll keep the intro short. I’ve got to update Promptcache with some new style nuggets. Enjoy the weekend!

---

News & Papers

Highlights

Midjourney Video

They did it, they finally did it.

This week Midjourney released the first version of their video model, and it has exceeded expectations. While the main specs don’t sound that special on paper (480p, 24fps, 5-second clips, extend feature), the magic sauce lies in their support for basically infinite animatable styles and especially in how incredibly easy it is to use thanks to their wonderful web experience.

Generating a video is as seamless as exploring different image generations and then simply hitting animate. Wait for about 60 seconds, and you’ll be greeted by 4 video variations. Like a variation? Hit extend and boom—a 9-second video after another 60 seconds. Super smooth. This works up to 4 times for a maximum video length of 20 seconds. I suspect it would be technically possible to extend infinitely, but storing longer videos would result in quite the hosting headache.

Yes, it’s heavily restricted when it comes to NSFW content. Yes, it’s still not super cheap. Yes, there’s a lack of control, and prompt adherence isn’t great yet. But my god, is it fun.

New Midjourney SREF System

As if a new video model wasn’t enough, Midjourney also released a new SREF system for V7. This means the old codes will not produce the same aesthetics as before, but you can still use them with --sv 4. I’ve updated Promptcache to reflect this change.

The new SREF system comes with better style variety and improved style adherence with image references. Compared to the old system, image references have much higher weight and are not so easily overridden by other influences like profile codes or moodboards. I’ve already started exploring new styles and will continue adding new favorite codes to the Promptcache collection.

3D

Hunyuan3D 2.1 can generate high-quality 3D assets from images through shape generation and texture synthesis.

PartPacker can generate high-quality 3D objects with many meaningful parts from a single image.

FreeTimeGS can reconstruct dynamic 3D scenes in real-time using Gaussian primitives that can appear at different times and places.

HOIDiNi can generate realistic human-object interactions with accurate hand contact and natural body movement from text prompts.

ImmerseGen can generate immersive 3D worlds from text prompts.

AnimateAnyMesh can animate 3D meshes based on text prompts.

Image

PosterCraft can generate high-quality aesthetic posters by improving how text and art work together.

MARBLE can blend and change the material properties of objects in images using material embeddings in CLIP-space. It allows control over attributes like roughness, metallic, transparency, and glow, enabling multiple edits at once and supporting various artistic styles.

UltraZoom can create gigapixel-resolution images from regular photos by upscaling them with detailed close-ups.

Video

ContentV can generate high-quality videos from text prompts in various resolutions and lengths.

UniRelight can relight single images or videos by estimating surface color (albedo) and creating realistic lighting effects in one step. It produces accurate shadows, reflections, and transparency.

DreamActor-H1 can generate high-quality videos showing how people use products, based on images of the person and the product.

PlayerOne can create dynamic 3D worlds from first-person images and generate videos that show real human motion.

InterActHuman can generate videos with multiple human characters by matching audio to each person.

MTV can create high-quality videos that match audio by separating it into speech, effects, and music tracks.

And that my fellow dreamers, concludes yet another AI Art weekly issue. If you like what I do, you can support me by:

• Sharing it 🙏❤️
• Following me on Twitter: @dreamingtulpa
• Buying me a coffee (I could seriously use it, putting these issues together takes me 8-12 hours every Friday 😅)
• Buying my Midjourney prompt collection on PROMPTCACHE 🚀
• Buying access to AI Art Weekly Premium 👑

Reply to this email if you have any feedback or ideas for this newsletter.

Thanks for reading and talk to you next week!

– dreamingtulpa