Hi, everyone! Here are the latest updates in the AI.
Hotshot
Hotshot started as a fun tool for making GIFs, but now it can also create full videos. Recently, they introduced a new feature that lets you make personalized videos by uploading up to 5 images and adding your own text prompts. You don’t even need to tweak the settings—just upload and go! Check it out here: Hotshot.
kohya_ss Update
The Kohya GUI, a tool for training AI models and creating LoRA (Lightweight Representations), just got a new update. Now, even if you have just a 6GB GPU, you can set up and train LoRA on Dream Booth FLUX Dev. It’s recommended to keep the old version while you try out the new one, just in case! https://github.com/bmaltais/kohya_ss/tree/sd3-flux.1
MeshyAI
Meet MeshyAI, a tool that can turn 2D images into 3D models in about a minute. It’s great for adding those hidden details that aren’t in the original image. Check it out at Meshy.
OmniCraft HDR Generator
OmniCraft is an AI tool that can make HDR (high dynamic range) lighting maps from both text and images. This means you can create realistic lighting effects for your projects. The release is just around the corner, so stay tuned!
Taaabs
Taaabs is a handy browser extension that works with a local AI model, meaning it doesn’t need to connect to the internet. It can save clips from web pages to your private library and even help you browse with an AI chatbot. Learn more about Taaabs on GitHub.
Qwen 2 VL 7B Sydney
This new vision language model, Qwen 2 VL 7B Sydney, processes images and text at the same time. It’s designed to give more positive, human-like responses, so it’s not just about facts—it’s about making the interaction feel natural and engaging. https://huggingface.co/adamo1139/Qwen2-VL-7B-Sydney
agentic_patterns
This one’s for the AI enthusiasts: agentic_patterns uses Groq LLM to give AI models more independence. It’s a design pattern that helps AI work on its own in specific tasks. You can find more on https://github.com/neural-maze/agentic_patterns
A Magic Prompt for Claude 3.5
There’s been a discovery in prompt design! With the right prompt, Claude 3.5 (an open-source AI) can be just as smart as some of the big-name models like OpenAI’s. It shows that with clever prompting, open-source models can achieve big things.
https://gist.github.com/philschmid/34747bf5bc8280f3a5f10f5fd8d1cd4b