100 relevant articles · classified by Haiku 4.5 · ingested daily
A community fine-tune of Krea 2 enables instruction-based identity-preserving image editing via a custom LoRA and ComfyUI node pack.
ComfyUI-Angelo adds support for Krea 2's Gen mode and Klein 9b Edit, plus fullscreen edits and outpainting.
A community developer shares an experimental ComfyUI workflow using Krea2 and Gemma 4 for cinematic storyboard generation.
Pallaidium demonstrates an open-source Blender workflow for converting 2D images to 3D scenes to talking video with camera control and scene consistency.
Google DeepMind adds background tasks and remote MCP support to Managed Agents in the Gemini API.
A legal battle over rights to the viral Skibidi Toilet series serves as a warning for digital creators and studios about IP ownership of AI-generated content.
Krea community contributor releases two new LoRAs for GPK and Ren & Stimpy styles, sharing training steps and dataset tooling.
A user reports improved character likeness and faster training with Krea 2, preferring it over Ideogram and Z Image for LoRA workflows.
SCAIL-2 segmentation control workflow released on CivitAI with image analyzer, LoRA support, and dual samplers.
Krea's V2 image generation model can interpret and replicate specific camera settings.
An independent developer previews a forthcoming open-source text-to-synth model with DAW-exportable audio and full metadata prompting.
Midjourney pushes Hollywood studios to disclose their AI usage, accusing them of hypocrisy.
A Star Trek fan film uses DaVinci Resolve, Fusion, and Unreal Engine for a 300-shot VFX pipeline.
A researcher releases SesquiLSR, a 3M-parameter learned latent upscaler for Flux, Anima, SDXL and other image models, as an open-source ComfyUI node.
Community release brings Qwen3.5 INT8 text encoders for ComfyUI, running 2B/4B/9B models on 8GB VRAM.
A developer shares an open-source tool that strips 90% of file size from Krea2 LoRAs by removing DIT/UNET weights while preserving text-conditioning layers.
Krea's v2 model demonstrates improved fine-art style fidelity in single-pass image generation without post-processing.
Developer releases a LoRA for Flux 2 Klein 9b enabling precise sun direction control in generated exterior images.
Community developer demonstrates character-consistent portrait generation on Krea2 using direct latent-space conditioning instead of text descriptions.
Lightricks launches the LTX Trainer with a community competition offering cash prizes and NVIDIA hardware for custom video LoRAs across five categories.
Burgstall Labs releases VR-Outpaint 1.0, an IC-LoRA for LTX2.3 that converts flat video into 360° equirectangular video via inverse gnomonic projection.
A user demonstrates using Krea2 with Comfy workflows to generate unlimited character-consistent image panels at full resolution for video production.
Scammers targeting ComfyUI extension developers with malicious npm packages to steal credentials and inject code into user extensions.
A community developer releases Local LLM Loader, a ComfyUI node for integrating local language models into image-generation workflows.
Krea 2 Turbo generates high-resolution images at scale with minimal user input.
Researchers propose Elastic Diffusion Transformer, an adaptive framework that accelerates diffusion models up to 2× while maintaining generation quality.
Together AI raises $800M at an $8.3B valuation, up from $3.3B in early 2025.
Next Lab Generation filmmakers were required to use Meta's Quill VR software and Magnific AI; most adopted VR but abandoned the AI tool.
Adobe acquires Topaz Labs, a developer of AI-powered photo and video editing tools.
KREA 2 generates complex image prompts in 15 seconds, demonstrating capability improvements since 2022.
Getty Images terminates its $3.7B acquisition of Shutterstock following UK regulator concerns over competition.
Community benchmark shows Krea2 INT8 ConvRot achieves significant speed improvements in ComfyUI 27.0 with comparable output quality.
ComfyUI MCP server enables AI agents to inspect, edit, and manage image-generation workflows programmatically.
A community member shares a Foley LoRA for LTX-2.3 on Hugging Face to generate synced sound effects without unwanted music.
U.S. Department of Commerce lifts export controls on Anthropic's Claude Fable 5 and Mythos 5 models.
Anthropic releases Claude Sonnet 5, a new iteration of its mid-tier LLM with improved performance.
User demonstrates Krea 2's image generation capability by reimagining The Matrix with 1960s era actors.
ComfyUI-Angelo adds outpainting and super-resolution upscaling features with UI improvements.
Developer demonstrates cinematic storyboard generation using Krea2 Turbo with Gemma 4, a custom node for panel splitting, and an optimized system prompt.
Krea2 model gains support in Forge Neo with strengths in fabric detail generation, though syntax-based prompt editing remains unsupported.
Together AI presents nine papers at ICML 2026 spanning inference, training, and model optimization.
Tidal demonetizes 100% AI-generated music effective immediately and will label them starting July 15th, while allowing them on the platform.
OpenAI and Work Louder are launching a hardware device with Codex shortcuts on July 15th.
EU regulators plan to require Google to allow third-party AI models system access on Android and share anonymized search data, prompting Google to warn of security risks.
Lawmakers propose expanding the Health and Location Data Protection Act to ban AI companies from selling user health and location data to brokers.
Comfy Org launches Comfy MCP in public beta, enabling AI agents to access ComfyUI workflows for image, video, 3D, and audio generation in natural language.
Zhipu AI's open-weight GLM-5.2 matches Anthropic's Mythos on cybersecurity tasks, narrowing the capability gap between Chinese and US models.
Krea2 generates high-quality images with advanced workflow techniques, gaining user adoption over competing tools like Ideogram.
Suno launches Spark, an incubator program offering grants and mentorship to independent musicians, with contentious licensing terms.
Nomad Studio releases 144 open-source cinematic image templates with Krea 2 prompts, each runnable in ComfyUI.
Clark Air releases a compressed 1.58-bit version of Stability AI's Sana 1.6B model achieving 8x compression with minimal quality loss.
ZPix adds support for Krea 2 Turbo with LoRA compatibility, generating 720p images in 40–54 seconds.
Krea 2 demonstrates strong performance on anime art generation, particularly in environment rendering compared to competing models.
A user releases a free PS2 framebuffer-style LoRA trained on Ideogram 4.0 with synthetic data, demonstrating style transfer across arbitrary scenes.
ChatGPT logs used as evidence in arson trial linked to deadly Palisades wildfire; prosecutors presented user's fire-related queries and image generation.
Krea releases Krea2, a style-transfer method for text-to-image generation tuned via modified RoPE technique.
Adobe acquires Topaz Labs, an AI image enhancement company, to integrate capabilities into Firefly and Photoshop.
Anthropic's Mythos 5 model regains partial access following Trump administration negotiations, with restrictions on public rollout of Fable 5.
A24 and Google DeepMind partner on a $75M deal to develop AI workflow tools for creative production, drawing backlash from filmmaker community.
Krea's 2-Turbo model generates high-fidelity behind-the-scenes and practical effects scenes with detailed prompt engineering examples.
Trump administration requests OpenAI delay GPT-5.6 rollout and limit early access to enterprise customers pending federal approval.
Krea 2 LoRA training produces comparable results to enhanced datasets with 50 images trained for 1,500 steps, though struggles with human body proportions.
ComfyUI adds INT8 quantization support, enabling faster inference with improved quality over FP8.
Community discusses SeFi-Image's semantic-first diffusion architecture and potential ComfyUI integration as Krea2 gains adoption.
Comfy Org releases Seedance 2.0 Mini and native 4K video generation in ComfyUI, expanding access to multi-shot AI video at lower cost.
Adobe acquires Topaz Labs, a specialist in AI-powered photo and video upscaling and enhancement tools.
Anthropic launches Claude Code in Slack with tagging; Figma and Google introduce new AI-powered design and browser control features.
Krea 2 demonstrates bounding-box prompting for spatial control in image generation, outperforming prior methods like Attention Couple.
Krea and Comfy founders discuss Krea 2 RAW, an unfinetuned base model licensed commercially below $1M revenue to support open-source innovation.
Krea 2 Turbo demonstrated generating 100+ art styles in a single scene for children's book illustration.
Community benchmark shows Krea 2 Turbo achieves 1.9× speedup with int8 quantization vs fp8 on Ampere GPUs, maintaining image quality.
HappyHorse 1.1 launches in ComfyUI with improved motion consistency, multi-character support, and synchronized audio generation for video production.
Rep. Luna's staff used Claude to draft defense-bill amendment summaries; congresswoman disputes broader AI use in legislation.
Krea 2 demonstrates improved cinematic image generation in community comparison with prior Z-image model.
User compares Ideogram 4 and Krea 2 image generation outputs, finding Krea 2 faster and Ideogram 4 stronger on fine details and composition.
Models accuse Rainbow Shops of using AI to generate lookalike images after their contracts ended.
User compares Krea 2 open-weight model performance against API versions and Ideogram 4 across prompt-following and image quality benchmarks.
Adobe launches its Firefly-powered AI Assistant across Creative Cloud, enabling agentic editing in Premiere, Photoshop, Illustrator, InDesign, and Frame.io.
Fastvideo releases FastWan-QAD, generating 5-second 480p video in 1.8 seconds on RTX 5090 with quantization-aware distillation.
A24 partners with Google DeepMind to apply AI to its film production workflow and creative processes.
Krea 2 open-source image models now run natively in ComfyUI with two checkpoints: RAW for fine-tuning and Turbo for fast 8-step inference.
Xindi Zhang's Oscar-shortlisted documentary animation 'The Song of Drifters' was built in ComfyUI, demonstrating the tool's capability for personal artistic vision.
Krea releases Krea 2, an open-source text-to-image model available in Raw and Turbo variants, ranked #1 on Artificial Analysis for independent lab models.
A custom node allows users to bypass Krea 2's built-in safety filters, circumventing the model's prompt-refusal mechanism.
Krea announces Krea 2 open-source release with GitHub, HuggingFace, and technical-report links.
Krea 2 Base and Turbo quantized models released free on HuggingFace in multiple formats (FP8, INT8, NVFP4) for ComfyUI across all GPU tiers.
Together AI releases ParallelKernelBench, evaluating LLM performance on writing optimized multi-GPU CUDA kernels across 87 workloads.
AI-generated virtual staging in real-estate listings misleads renters with fabricated apartment conditions, creating costly false expectations.
A24 and Google DeepMind partner on a $75M research collaboration with no production mandate.
Google DeepMind partners with A24 on AI-driven movie production tools, investing around $75M in the film studio.
Tesla Autopilot under NHTSA investigation after a Model 3 crashed into a home at high speed, killing an elderly woman.
Comfy Org ships Comfy Go, a native SwiftUI mobile app for image and video generation on Comfy Cloud, demonstrating API accessibility.
Oak, a Git alternative designed for AI agents, launches with virtual mounts to reduce repository overhead and enable parallel task execution.
OpenAI and Getty Images forge a multi-year partnership integrating licensed Getty content into ChatGPT's search and discovery features.
An open-source LoRA for music generation trained on dark-electronic instrumental tracks is shared on Hugging Face.
Developer releases an MIT-licensed desktop app for building storyboards with Ideogram 4 via ComfyUI, featuring structured prompt editing and LLM integration.
