ComfyUI ecosystem updates — open-source model releases, custom nodes, workflows, and tools for image, video, and audio generation.
Text to imageM87 is an early-preview aesthetic LoRA for Krea 2 Turbo that enhances composition, lighting, atmosphere, and visual texture — making generations feel more cinematic and art-directed without locking into a specific style.
Alissonerdx releases LTX-Best-Face-ID, a reference-to-video identity LoRA for LTX-2 using overlap conditioning, TASS-RoPE source-phase tagging, and an ArcFace identity loss. No white-background reference needed in ComfyUI.
Image editingOstris pushed support in AI Toolkit for training Krea 2 models with reference images, enabling edit-style LoRA training that masters concepts like 'make this a cyclops' in just 1,750 steps.
MIT
Text to imageTanmay Patil released the Krea 2 Depth ControlNet-LoRA, adding depth-conditioned image generation to Krea 2. Works with both Krea 2 Raw and Turbo, with depth consistency reaching 0.99 Pearson correlation.
Krea-2-Community
Text to imageFlux2-Klein-9B-True-V2 is the latest full fine-tune by wikeeyang based on FLUX.2-klein-9B, delivering significant improvements in photorealism, prompt adherence, LoRA compatibility, and image editing quality.
FLUX-1-dev-non-commercial
EcosystemMeituan releases LongCat 2.0, a 1.6 trillion parameter MoE language model with 1 million token context window, trained entirely on AI ASIC hardware. Model weights coming soon under MIT license.
Video editingAlissonerdx releases Edit Anything, an Apache-2.0 open-source video editing LoRA for LTX-2.3, featuring motion transfer, no-reference multitask editing, and reference video-to-video — all through ComfyUI BFSnodes.
Comfy Org launches Comfy MCP, connecting AI agents to ComfyUI via the Model Context Protocol. Agents can now search models, execute workflows, and generate images, video, and 3D content in natural language.
MultimodalKandinsky Lab (Sber) releases KVAE-Audio, a 166.9M parameter continuous audio autoencoder achieving state-of-the-art reconstruction quality across speech, music, and general sound at 48kHz full bandwidth.
ilkerzgi has released over 1,500 Krea 2 Style LoRAs on Hugging Face, spanning 7 categories from illustration to photographic. Each LoRA comes ready to use in ComfyUI.
Video editingfal.ai released LTX-2.3 3DREAL IC-LoRA, an in-context LoRA for LTX-Video that transforms rough 3D blockouts and CG renders into photorealistic, cinematic video while preserving the original composition and camera motion.
Open-Weights
MultimodalNVIDIA open-sources LocateAnything-3B, a vision-language grounding model featuring Parallel Box Decoding (PBD) for fast and precise object localization, supporting object detection, GUI element grounding, OCR localization, and point-based grounding across diverse domains
Open-Weights
Video editingWhatDreamsCost releases LTX Director 2.0, a massive update to the free open-source ComfyUI video editing tool adding complete video support, IC-LoRA integration, audio inpainting, retake mode for selective regeneration, and a full UI overhaul.
Text to videoFuzzPuppy releases LTX-2.3 Foley LoRA, a video-to-audio LoRA that adds realistic, visually synchronized sound effects to LTX-2.3 generated videos without background music overlay.
Image to videoHKUST C4G releases DomainShuttle, an Apache-2.0 open-domain subject-driven video generation model built on Wan2.2-T2V-14B. Features Domain-MoT, Video-Reference DualRoPE, and Cross-Pair Consistent Loss for flexible in-domain fidelity and cross-domain style transfer.
Apache-2.0
Text to imageKrea.ai released Krea 2 Raw and Turbo, an open-weight 12.9B parameter Diffusion Transformer for text to image generation. ComfyUI supports it natively with ready to use workflows.
Krea-2-Community
Image editingBoogu-Image-0.1-Edit is an Apache 2.0 licensed image editing model from the Boogu-Image family, offering instruction-based image editing with a unified multimodal understanding and generation architecture.
Apache-2.0
Text to speechHiggs TTS 3 is a 4B parameter text-to-speech model supporting 100+ languages with zero-shot voice cloning, expressive emotional control, and inline prosody/sound effects for voice agent applications.
Open-Weights
Text to videoSulphur 2 is a community fine-tune of LTX 2.3 offering text-to-video and image-to-video generation with a built-in prompt enhancer and distill LoRA, trained on 125K+ curated clips.
Open-Weights
OpenMOSS team releases MOVA (MOSS Video and Audio), an end-to-end synchronized video and audio generation foundation model that generates video and audio in a single inference pass, achieving precise lip-sync and environment-aware sound effects, fully open-sourcing model weights, training and inference code
Open-Weights