All updates

Can MiniMax H3 keep up with video playback on one GPU while still producing clips worth watching? ComfyStreamerH3 is my attempt to find out.

A 17-minute Backrooms short film, made by one person with AI. Javi Lopez, co-founder of Magnific, walks me through how he pulled it off, the "don't do handheld" lesson, and the thing no one tells you about making a film this long alone: form a band. This is the first of what…

AI coding agents can now mod almost any PC game you own, so I put DOOM inside Skyrim. SkyDoom is real Skyrim with all the NPCs, except Doom monsters spawn everywhere and I have a chaingun. Built with UniversalModder in about an hour, and I'll show you how to mash up any two…

Text and image to video · 1–15 s · 480p, 720p, 1080p · 16:9 9:16 1:1 4:3 3:4 3:2 2:3 · from $0.020/s on OpenRouter

In this tutorial, I'll share with you one of my favorite Free AI Tools. You can download it and run it locally on your Mac (woot!), or you can use Fal.ai. Using a depth map as a reference is a phenomenal way to ensure your AI videos have strong accuracy when changing…

In this video, we tested some of the LAtest AI tools to see how they are affecting sound design and the future of video production. We tested multiple workflows against a Curious Refuge member's work and came to a conclusion on the current state of AI sound design tools. Check…

Four is your lucky number today. Kling 4 is here — kinda. But, I ran it through the usual tests: text-to-video, image-to-video, Omni with Flamethrower Girl, a noir scene in Malloy's office, and my keyframes-plus-Omni trick. Then Google wakes up with Gemini 4 Argon (announced…

The behind-the-scenes video was made to be part of the story.

In this video, we test all of the top AI Creative Agents against the work from a real AI artist to see how AI holds up when compared to Human creativity. We also dive into some new AI Filmmaking tech that might begin to make AI Creative more obsolete in the near future. Check…

Text and image to video · 5–30 s · 480p, 768p, 1080p, 2K · 21:9 16:9 4:3 1:1 3:4 9:16 · from $0.020/s on OpenRouter

In this tutorial, we teamed up with our AI Animation Instructor, @albertbozesan to show you the latest AI workflow for combining 2D animation with live-action footage. Check out a full breakdown here: https://curiousrefuge.com/blog/the-best-ai-workflow-for-a-2d-live-action-film…

In this video, we let GPT Astra organize and edit a video including doing VFX, SFX, and producing marketing assets. Check out our tutorial for what worked and what didn’t. Check out the full written breakdown and prompts here…

Learn how to use MiniMax Fast H3 in ComfyUI with three video workflows: text to video, first frame to video, and first and last frame to video. I’ll show you how to set up the model, choose your video size and duration, create prompts with Google Gemini, and run each workflow…

In this video, we tested a hybrid filmmaking workflow that combines live-action footage with AI to transform simple greybox shots into entirely new environments. Check out the full written breakdown and prompts here…

The MiniMax H3 video VAE now encodes up to ~2.2x faster and decodes ~1.4-2.7x faster on Nvidia GPU.

In this video, we tested three different ways to create new camera angles with AI, and the results were pretty wild. Check out the full written breakdown and prompts here: https://curiousrefuge.com/blog/how-to-change-camera-angles-with-ai Check out our courses…

H3 Max generates a 5-second video in under 3 seconds. In our human preference evaluations against twelve leading video models, it ranks first on quality, prompt understanding, and aesthetics. Independent benchmarks from Artificial Analysis and Design Arena rank it first as well…

Running MiniMax H3 on your own computer at home is great, but generating higher resolution videos can take a while. Thankfully we have various upscale methods to help us! Using Ultimate Upscale, Latent Upscale 3D and SeedVR2 you can quickly generate your lower resolution (0.2…

World Labs Atlas is a new omni world model for spatial intelligence from Fei-Fei Li’s company. It generates, reconstructs, and simulates 3D worlds from text, images, video, and camera poses, not just flat clips. Atlas claims pixel-perfect camera control, up to 1 minute of 1440p…

Another ComfyUI workflow power-up to dive further into the world of extending pre-existing or freshly generated videos using the MiniMax H3 Reference conditioning nodes, testing Sparse Linear Attention (SLA) along with the new Lightx2v SLA LoRA and the PDD Acc LoRAs too…

Google's updated Omni Flash model brings 4K output, keyframe interpolation, and conversational video editing to your workflow.

Thanks for reading ComfyUI Newsletter!

Here is how to create interesting, deep, rich ART with AI. It is a lesson I have learned over many years of studying and observing art. In this video, I will show you visual examples and break it down for you in a really simple way to give you the perfect starting point to…

Announcing H3 Max, a post-trained version of MiniMax H3 developed by fal Research and optimized for maximum speed by fal's inference team. In our human preference evaluations, H3 Max ranks #1 across overall quality, prompt understanding, and aesthetics against leading video…

Speed up MiniMax H3 video generation in ComfyUI with the Speed LoRA, optimized sampler settings, and Pixaroma VRAM monitoring nodes. In this tutorial, I show you how to reduce MiniMax H3 generation from the standard 20 steps to 8 or even 4 steps, while finding a practical…

Alibaba's new video generation model brings 30-second single-pass generation, up to 20 reference assets, and instruction-driven video editing to your workflow.

Most people treat the Dead Internet Theory as a death sentence for the web. Here is a different perspective In this video I explore why the rise of agentic AI bots, especially ones that pay per crawl, could actually be the beginning of something better. Instead of a lifeless…