DeepSeek-V4-Flash Makes LLM Steering Practical
DeepSeek-V4-Flash is the first local model competitive with frontier AI, making LLM activation steering practical for the first time. A guide for creators.
SANA-WM: Open-Source AI Generates 1-Minute 720p Video
NVIDIA's SANA-WM is a 2.6 billion-parameter open-source world model that generates 720p, 60-second video with 6-degree-of-freedom camera control on a single GPU.
ComfyUI-DramaBox Adds LoRA Support for Custom Voice Training
ComfyUI-DramaBox added LoRA weight injection on May 16, letting creators load custom voice personalities into workflows without model reloads.
Supertonic 3: On-Device TTS Speaks 31 Languages
Supertonic 3 is an open-weights, CPU-only TTS engine from Supertone with 31 languages, expression tags, and zero-shot voice cloning.
FLUX.2-klein-9B in ComfyUI: Sub-Second Image Gen
Black Forest Labs' FLUX.2-klein-9B now integrates with ComfyUI via NKD Klein Tools v1.7.0, enabling sub-second image generation with just 4 inference steps on an RTX 4090.
holaOS 0.1: AI Workstream Manager for Creators
holaOS 0.1 ships Dashboard, Sub Agents, and Multi Workspaces for managing parallel AI workflows on macOS.
Inside Bun's 1M-Line Rust Rewrite by Claude Code
Anthropic's Claude Code ported Bun from Zig to Rust in nine days: 1,009,257 lines, 99.8 percent test pass, 13,000-plus unsafe blocks. The case study.
IBM Granite Embedding R2: Open Multilingual RAG Models
IBM has released Granite Embedding Multilingual R2: a pair of Apache 2.0 embedding models with 32K context, 200+ languages, and a top MTEB score under 100M parameters. A drop-in swap for paid commercial embedding APIs.
VOID, BiRefNet, and Gemma 4 Are Now in ComfyUI
Three significant open-source models arrived in ComfyUI on May 14, 2026: VOID for video object deletion, BiRefNet for image segmentation, and Gemma 4 for multimodal reasoning.
Image-Blaster: One Image to 3D Worlds for Claude Code
Open-source Claude Code skill turns one image into a fully meshed 3D scene with separate object meshes, Gaussian splat background, and ambient plus physics audio in under five minutes.
LTX Director v1.3.0: ComfyUI Timeline Editor for LTX 2.3
LTX Director v1.3.0 dropped today: a single ComfyUI node that replaces your entire LTX 2.3 toolkit. Timeline editing, prompt relay, 8 keyframes, and custom audio, all in one place.
OpenReader v3: Self-Hosted TTS for PDF, EPUB, DOCX
OpenReader v3.0 converts PDF, EPUB, DOCX, TXT, and Markdown files into synchronized read-along sessions or exported audiobooks, with multiple TTS providers and Docker deployment.
Scenema Audio: Open-Source Voice Cloning From LTX 2.3
ScenemaAI released Scenema Audio on Hugging Face and GitHub, an open-weights expressive TTS and zero-shot voice cloning model built on the audio half of Lightricks LTX-2. MIT inference code, 13 languages, real-time on a 24 GB GPU.
Cline Releases SDK to Embed Coding Agent Anywhere
Cline released @cline/sdk on May 13, 2026, an open-source TypeScript runtime that lets developers embed the same coding agent powering Cline.
Hermes Agent: Self-Improving AI for NVIDIA RTX Creators
Nous Research's Hermes Agent reached 140,000 GitHub stars in three months. NVIDIA featured it for RTX creators who want a local AI agent that writes its own skills.
OpenSquilla v0.1: Cut Agent Token Costs 60% to 80%
OpenSquilla v0.1.0 launched May 12, 2026: an Apache 2.0 microkernel runtime that smart-routes agent traffic across model tiers for 60 to 80 percent token savings.
Pixal3D: Open Image-to-3D Model Accepted to SIGGRAPH 2026
TencentARC Pixal3D generates 3D models that stay pixel-aligned with your input image. Accepted to SIGGRAPH 2026, with a live HuggingFace demo and open-source code.
MiniCPM-V 4.6: 1.3B Open Vision Model Runs on Phones
OpenBMB shipped MiniCPM-V 4.6, a 1.3B Apache 2.0 vision-language model that runs on iPhone, Android, and HarmonyOS and leads the sub-2B open-weights pack.
HiDream O1: Open-Source 8B Image Model Beats FLUX at 2K
HiDream-O1-Image is an open-source 8B pixel transformer that beats FLUX.2 Dev at 2K resolution, with MIT license and ComfyUI support.
Moonshot AI Hits 20B as Open-Weights Funding Surges
Moonshot AI closed a 2 billion dollar round at a 20 billion dollar valuation led by Meituan May 7. What the open-weights tier consolidation means for creators.
Best ComfyUI Workflows 2026: 12 Production-Ready Setups
By mid-2026 ComfyUI hosts thousands of community workflows. Here are 12 that have crossed the production-quality threshold, with download links and the use case each one actually serves.
IBM Granite 4.1: Dense LLMs Walk Back the MoE Bet
IBM released Granite 4.1 on April 29, 2026 with dense decoder-only 3B, 8B, and 30B Apache 2.0 LLMs that walk back the Granite 4.0 MoE bet. The 8B dense beats the 32B-A9B MoE on most benchmarks. 128K production context, 512K extension.
Zed 1.0 Lands With Multi-Agent Coding and Edit Prediction
Zed Industries shipped Zed 1.0 on April 29 2026, a Rust-built, GPU-accelerated code editor with parallel agents, edit prediction, and the open Agent Client Protocol.
Mistral Medium 3.5: 128B Open Weights, Cloud Vibe Agents
Mistral shipped Medium 3.5 as a 128B dense multimodal model under modified MIT, with cloud Vibe agents and Le Chat Work mode on day one.