We fixed the same ComfyUI errors enough times to write them all down. 30+ errors. Security blocks, import failures, VRAM crashes, missing models, video gen limits. SageAttention "wrong Python" fix, VRAM tables by model and GPU tier, the full model folder structure, Wan 2.2 frame limits. One page, every fix, with screenshots. Full guide below ⬇️

Ming Image 0.1 Design. 6B params. Open source. MIT license. Fashion → retro editorial → animated styles → aerial photography. One model. Built-in prompt enhancer. Transparent PNG output. Up to 2048px. Try it on Floyo 👇

Swap the entire face in a portrait with Qwen Image 2.1. Hair, jawline, and head shape included, while keeping the pose and background intact. @Alibaba_Qwen Create now ⬇️

One image in. 60 FPS anime video with native audio out. KatoriKonoe built a single-queue workflow on Floyo: Illustrious SDXL → MiniMax H3 → RIFE interpolation → audio VAE. No post-processing, no separate audio sync. Character motion holds. Audio matches the scene. @MiniMax_AI Try it now ⬇️

AI art moved fast this week. 6 updates worth knowing: 1. Nano Banana 2.1 dropped. 14-reference fusion, text rendering that actually works, up to 4K. Google's quiet model just got loud. 2. ChatGPT is bringing ads into image generation. They show up while your image renders. 3. Tencent Hy Image 3.5 Preview. $0.024/image with ComfyUI nodes. 4. Seedream 5.0 Flash. $0.018/image, 15 seconds. A year ago this cost 10x more. 5. Kling 4.0 Flash in early access. Up to 4K, stereo audio, 10-bit HDR. 6. Vidu Q4 Preview just dropped. 15 image refs, 3 audio refs, 4K. Built for production. Better models. Bigger possibilities. Smarter workflows. Full info in replies. ⬇️ @Google, @OpenAI, @TencentHunyuan, @BytedanceTalk

Two images are all you need for a virtual try-on. Qwen Image 2.1 swaps the outfit while keeping the person’s face, pose, and background intact. Useful for testing different looks, product variations, or clothing concepts, with support for up to 10 reference images. Workflow below ⬇️ @Alibaba_Qwen

Wan Animate 2 vs Wan 2.1 + SCAIL-2 vs MiniMax H3 for motion transfer. Same source video. Same character. Completely different strengths. Which one would you pick? Full comparison in the comments. ⬇️ @Alibaba_Wan, @Zai_org, @MiniMax_AI

Loved seeing so many creators come together! 100+ creators joined us for a live workshop with Luka Tisler, diving into MiniMax H3. From LoRAs and ControlNet to frame control and prompting, he covered what works, what doesn’t, and how to get the most out of H3. Real demos, practical tips, and techniques you can put to use. A big thank you to Luka and everyone who joined us and made this such a wonderful experience! The recording and workflow are now live. Link in comments ⬇️