Lumina-T2X: Transforming Text into Any Modality, Resolution, and Duration via Flow-based Large Diffusion Transformers abs: arxiv.org/abs/2405.05945 code: github.com/Alpha-VLLM/Lum… Introduces the Lumina-T2X family of models, with the largest being a 7B DiT. Able to generate images,

Vchitect-2.0 new text and image to video model Parallel Transformer for Scaling Up Video Diffusion Models

Thanks AK for sharing! 🎉 🔥🔥🔥We've released Vchitect-2.0, a 2B video generation model, supporting up to 720x480 resolution and 5-20 second generation. 👉 Website: vchitect.intern-ai.org.cn 👉 Code: github.com/Vchitect/Vchit… 👉 Demos Hugging Face: huggingface.co/spaces/Vchitec…

📢High-Quality Text-to-Video (T2V) Dataset📢 We are now releasing 🎬Vchitect-T2V-Dataverse🎬, a large-scale T2V database with *2M high-quality video clips* as well as *fine-grained textual captions* - Code: github.com/Vchitect/Vchit… - Dataset Hugging Face: huggingface.co/datasets/Vchit…

🔥Time to Upgrade Your Classifier-Free Guidance🔥 🌠CFG-Zero*🌠 offers consistently better *visual quality* and *text alignment* on text-to-image/video - Project: weichenfan.github.io/webpage-cfg-ze… - Code: github.com/WeichenFan/CFG… - Demo Gradio: huggingface.co/spaces/weepies… . Thanks AK!

While the world is clamoring for the new OpenAI 4o image generation model, the open AI community makes yet another step towards SOTA image quality with CFG Zero Star technique 😱 Now added to the upcoming APW 13.0 EA2 for ComfyUI.

🔥🔥🔥WaN 2.1-14B T2V is now officially supported in our repo! 🎉 With 4% steps of zero-init, frame quality sees a significant boost. 🔗 github.com/WeichenFan/CFG…

FLUX.1 Dev and SD3.5 are not the only ones to benefit from the new CFG Zero Star technique. Even WanVideo 2.1 benefits from it. APW 13.0 EA2 for ComfyUI will support this use case, too.

🔥🔥🔥 We now officially support wan2.1-14B for Image-to-Video generation! Check out the latest update in our repo: 👉 github.com/WeichenFan/CFG…

Our method ✨CFG-Zero*✨works well with LoRA too! Generated using Flux with the 👻Death Stranding👻 LoRA applied. 👉Check out our repo: github.com/WeichenFan/CFG… LoRA: civitai.com/models/46080?m… #DeathStranding #VirtualPhotography #KojimaProductions

🌟 CFG-Zero* now supports Qwen2.5-Omni! 🔗 github.com/QwenLM/Qwen2.5… Big thanks to Wan2.1 for the mention! 🔗 github.com/Wan-Video/Wan2… Check out our latest updates: 👉 github.com/WeichenFan/CFG… 🎧 Generated audio samples below ⬇️

✨ CFG-Zero★ is now supported in 🤗 Diffusers(github.com/huggingface/di…)! Use it with just one line: "guidance = ClassifierFreeZeroStarGuidance(guidance_scale=5.0, zero_init_steps=1)" 📦 Repo: github.com/WeichenFan/CFG… The comparisons produced by Diffusers:

✨CFG-Zero* supports EasyControl(github.com/Xiaojiu-z/Easy… Jiaming Liu) now! 📷Check our repo for the lastest update on Ghibli-style generation: github.com/WeichenFan/CFG… #ghiblistyle

CFG Zero star has been integrated into the EasyControl-Ghibli space! Try it out and have fun. huggingface.co/spaces/jamesli… Thanks Weichen FAN Ziwei Liu

🎬 𝗖𝗩𝗣𝗥 𝟮𝟬𝟮𝟱 𝗧𝘂𝘁𝗼𝗿𝗶𝗮𝗹 𝙁𝙧𝙤𝙢 𝙑𝙞𝙙𝙚𝙤 𝙂𝙚𝙣𝙚𝙧𝙖𝙩𝙞𝙤𝙣 𝙩𝙤 𝙒𝙤𝙧𝙡𝙙 𝙈𝙤𝙙𝙚𝙡 🚀 Hosted by MMLab@NTU × Kuaishou, etc 📅 June 11 | Nashville 🔗 world-model-tutorial.github.io 🧠 Video is just the start. World modeling is the goal. #CVPR2025 #WorldModel

🎬#CVPR2025 𝐓𝐮𝐭𝐨𝐫𝐢𝐚𝐥 🗺️𝑭𝒓𝒐𝒎 𝑽𝒊𝒅𝒆𝒐 𝑮𝒆𝒏𝒆𝒓𝒂𝒕𝒊𝒐𝒏 𝒕𝒐 𝑾𝒐𝒓𝒍𝒅 𝑴𝒐𝒅𝒆𝒍 #CVPR2025 🔗world-model-tutorial.github.io 📅June 11 🚀Hosted by MMLab@NTU x Kling AI 🧠Incredible lineup of speakers: Jack Parker-Holder Hong-Xing "Koven" Yu Jiaming Song Pengfei Wan Angjoo Kanazawa Sherry Yang

#Genie3 Awesome release by Google DeepMind! We finally reach an incredible moment for interactive world models! Please also checkout the recording and slides of our tutorial on #CVP2025 where Jack Parker-Holder shared his thoughts on scaling world models! world-model-tutorial.github.io