ComfyUI News & Open-Source AI Releases

Microsoft Asia releases Mage-Flow -- compact 4B model for text-to-image and editing, native resolution up to 2048, RL/4-step Turbo variants, MIT license.

MIT

Anima Control Pose Preview-2 brings pose-controlled image generation to Anima with native ComfyUI custom nodes, multi-resolution support at 512/768/1024, and improved pose accuracy

Lightricks releases Clean Plate IC-LoRA for LTX-2.3 — a video-to-video LoRA that removes people and objects from footage and reconstructs the empty background, no mask required.

tarn59 released two style transfer LoRAs for Bernini-R — anime style and real life style — enabling video-to-video style conversion directly in ComfyUI.

Apache-2.0

DreamFast releases an uncensored Qwen3-VL-4B text encoder for ComfyUI, achieving 100% HarmBench ASR across BF16, FP8, INT8, NVFP4 and MXFP8 formats.

Apache-2.0

Lightricks releases an official LTX-2.3 Foley V2A LoRA for realistic, visually synchronized sound effects — footsteps, impacts, typing, and environmental audio.

ComfyUI v0.28.0 adds native SeedVR2 upscaling, NVIDIA PID 1.5 PixelDiT support, convrot int4 optimizations, new Save 3D and Text Overlay nodes, and partner node updates including sync.so sync-3 and Seedream 5 Pro.

Tongyi Lab (Alibaba) releases Wan-Dancer-14B, an open-source model that generates 720p/30fps dance videos from music and reference images, supporting Chinese Classical, K-Pop, Street, Tap, and Latin dance genres for minute-long outputs.

Fal releases Ideogram V4 Fast (20-step FP4, 5s inference) and Ideogram V4 Instant (8-step BF16, 2s inference), speed-distilled variants with no-CFG support. Available in int8-convrot format for ComfyUI via Hippotes.

Open-Weights

Cseti releases a novel IC-LoRA for LTX-Video 2.3 that acts as a virtual second camera, letting users re-render video scenes from new camera angles using simple text prompts.

Apache-2.0

Cseti releases a proof-of-concept IC-LoRA for LTX-2.3 that acts as a virtual second camera, letting you change the camera angle of existing video footage using a discrete prompt vocabulary.

LiconStudio releases V2 of the popular Multiple Subject Reference LoRA for LTX-2.3, with improved consistency, stability, and scene logic based on community feedback.

Alaya Lab released AlayaWorld, a full-stack open source framework for building interactive generative worlds. It features long-horizon video consistency beyond one minute, real-time camera control, and prompt-driven interaction.

rzgar releases Bernini-R-S2V, a speech-to-video fine-tune that adds Wan2.2 S2V audio-driven lip-sync capabilities to ByteDance's Bernini-R model, with dedicated ComfyUI custom nodes and multiple precision variants.

Apache-2.0

Ant Group's Robbyant open-sources LingBot-Video, a 30B-A3B MoE video generation model achieving SOTA on RBench with Apache 2.0 license.

Apache-2.0

A new LoRA from ostris enables style reference transfer on Krea 2 Turbo, trained on thousands of hand-curated style pairs with AI Toolkit.

Lightricks releases the Cinemagraph IC-LoRA for LTX-2.3, an open-weight adapter that keeps backgrounds frozen while moving only a single element — delivering seamless, natural loops from any still image.

LTX-2 Community License

SenseTime releases SenseNova-Vision-7B-MoT, a 7B-parameter model that unifies detection, segmentation, depth estimation, normal prediction, and multi-view geometry into a single instruction-following framework — no task-specific heads needed.

CC BY-NC 4.0

Anima, the 2B anime text-to-image model by CircleStone Labs, gets two new official variants: a fast Turbo version and a quality-tuned Aesthetic version.

circlestone-labs-non-commercial-license

ComfyUI-Angelo now supports two-model workflows, letting users generate with Krea 2 (or any other model) and edit with FLUX 2 Klein 9B in a single unified canvas — no manual graph rewiring needed.