AI Art Weekly #95
Hello there, my fellow dreamers, and welcome to issue #95 of AI Art Weekly! 👋
Mini-Tulpa is taking its fair time to arrive, so I still have some to write this newsletter for you 🫡
Also added another 20+ SREF codes and a bunch of prompts to PROMPTCACHE this week, we’re now at 60+ codes and 750+ prompts. If you’re looking for some fresh inspiration, you know where to find it!
In this issue:
- Highlights: Grok 2 + FLUX.1, Midjourney Web Editor
- 3D: Photometric Inverse Rendering, HeadGAP, Sketch2Scene
- Image: MagicFace, Generative Photomontage, UniPortrait
- Video: Puppet-Master, Kalman-Inspired Feature Propagation
- and more!
Unlock the full potential of AI-generated art with my curated collection of Midjourney SREF codes and prompts. The first 20 can use AIARTWEEKLY during checkout to get $10 off!
Cover Challenge 🎨
For the next cover I’m looking for submissions within submissions! Reward is again fame & glory and a rare role in our Discord community which lets you vote in the finals. Rulebook can be found here and images can be submitted here.
News & Papers
Highlights
Grok 2 + FLUX.1
xAI made a lot of waves this week by releasing Grok-2 and Grok-2 mini, their latest language models with advanced reasoning capabilities. Key features include:
- Improved performance in chat, coding, and reasoning tasks
- Can generate images with FLUX.1
- Outperforms Claude 3.5 Sonnet and GPT-4-Turbo on the LMSYS leaderboard
- Excels in vision-based tasks, achieving state-of-the-art performance in visual math reasoning and document-based question answering
- Enhanced capabilities in following instructions and providing accurate information
- Significant improvements over the previous Grok-1.5 model in various academic benchmarks
The models are currently available in beta on 𝕏 for Premium users.
I personally played around with image generation and the speed, quality and less restrictive nature of this implementation make it a lot of fun. Because FLUX can create almost perfect photorealistic fake images, not everybody is happy that this tech is now so easily accessible by the masses. Fact is though, that it’s here and it’s here to stay. We have to accept the fact that everything we see on a screen can be fake. The sooner we learn this, the better.
low quality cctv footage of a woman standing at the corner of a busy street holding up a cardboard with the phrase "WILL DO ANYTHING FOR PROMPTCACHE ACCESS".
Midjourney Web Editor
Midjourney introduces a new web editor, streamlining image manipulation into a single, unified interface. This now makes it possible to inpaint and outpaint images in one go. Especially the unrestricted outpainting enables a lot of new creative possibilities.
Midjourney Web Editor in action
3D
Photometric Inverse Rendering: Shading Cues Modeling and Surface Reflectance Regularization
Photometric Inverse Rendering can figure out light positions and reflections in images, including tricky shadows. The method it employs breaks down surface reflections better than other tools, working well on both fake and real pictures.
PIR examples
HeadGAP: Few-shot 3D Head Avatar via Generalizable Gaussian Priors
HeadGAP can create lifelike 3D head avatars from just a few photos. It uses Gaussian Splatting to make heads that look real from different angles and can be animated smoothly.
HeadGAP examples
Sketch2Scene: Automatic Generation of Interactive 3D Game Scenes from User’s Casual Sketches
Sketch2Scene can create interactive 3D game scenes from simple sketches and text descriptions. It uses a diffusion model with ControlNet and procedural generation to make high-quality, playable 3D environments that match what users want.
Sketch2Scene example
Image
MagicFace: Training-free Universal-Style Human Image Customized Synthesis
MagicFace can generate high-quality images of people in any style without needing training. It uses special attention methods for precise attribute alignment and feature injection, working for both single and multi-concept customization.
MagicFace examples
Generative Photomontage
Generative Photomontage can combine parts of multiple AI-generated images using a brush tool. It enables the creation of new appearance combinations, correct shapes and artifacts, and improve prompt alignment, outperforming existing image blending methods.
Generative Photomontage examples
UniPortrait: A Unified Framework for Identity-Preserving Single- and Multi-Human Image Personalization
UniPortrait can customize images of one or more people with high quality. It allows for detailed face editing and uses free-form text descriptions to guide changes.
UniPortrait examples
Video
Puppet-Master: Scaling Interactive Video Generation as a Motion Prior for Part-Level Dynamics
Puppet-Master can create realistic motion in videos from a single image using simple drag controls. It uses a fine-tuned video diffusion model and all-to-first attention method to make high-quality videos.
Puppet-Master examples
Kalman-Inspired Feature Propagation for Video Face Super-Resolution
KEEP can enhance video face super-resolution by maintaining consistency across frames. It uses Kalman filtering to improve facial details, working well on both synthetic and real-world videos.
KEEP example
Audio
PeriodWave: Multi-Period Flow Matching for High-Fidelity Waveform Generation
PeriodWave can generate high-quality speech waveforms by capturing repeating sound patterns. It uses a period-aware flow matching estimator to outperform other models in text-to-speech tasks and Mel-spectrogram reconstruction.
Zero-shot text-to-speech results with PeriodWave
Also interesting
@bengttibert, @nathanboey, and @mind_wank put together this incredible 11 minute pilot of their new show AI OR DIE. A must watch.
@iamneubert from Runway shared his GEN-3 image-to-video prompting process. Always interesting to see the workflows of others.
If you’re looking to train your own FLUX LoRAs, @HBCoop_ put together an expansive thread of resources on how and where to do this. I might release something myself in the upcoming weeks 👀
@EccentrismArt took Gen-3’s impressive emotion generations and added LivePortrait on top of it. What will this look like in a year? 🤯
@HBCoop_ put together a super dope 8-Step workflow and tutorial below which she used to create a cosmetic ad with AI.
@8bit_e created this fun short AI animation by feeding his movements to his Comfy workflow. Sound effects were created with ElevenLabs.
“Find me at Midnight” by me.
And that my fellow dreamers, concludes yet another AI Art weekly issue. Please consider supporting this newsletter 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 😅)
- Buy my Midjourney prompt collection on PROMPTCACHE 🚀
Reply to this email if you have any feedback or ideas for this newsletter.
Thanks for reading and talk to you next week!
– dreamingtulpa