#Image generation

AIMidjourney has updated its V8.2 edit model to improve image quality. Users who hit problems in the past 24 hours are invited to retry and keep sending feedback, with more updates promised soon.

AIRadixArk introduced LoRA SFT in Miles-diffusion for fast, targeted post-training of diffusion models. The company trained a rank-64 LoRA adapter for MiniMax H3 to improve physical realism, using 254 curated training windows and under 3 hours on 8 GPUs. The adapter can be exported to safetensors and served directly with SGLang without retraining the full model.

AINVIDIA's DLSS 4.5 Ray Reconstruction is now available, using a second-generation joint denoiser and super-resolution model. According to the post, it delivers much better image quality at the same compute cost, pushing the trade-off between image quality and rendering cost further.

AIMicrosoft AI introduced MAI-Image-2.6, its latest image model, which can modify color, style, and visual details while maintaining consistency across iterations. The post positions it as a tool for bringing mood boards to life.

AIGoogle has published TIPS B/14 (v1) on Hugging Face, a contrastive vision-language model that produces spatially rich image features aligned with text embeddings. The model has 86M vision parameters and 110M text parameters at native 448 resolution, and is licensed under Apache 2.0. The release includes usage code for image and text encoding, zero-shot classification, and spatial feature visualization.

AIGoogle shows Gemini 3.7 Flash, paired with Nano Banana and Omni, generating a fully animated parallax landing page from a single prompt. The model calls tools to write the copy, create image references, and render parallax-ready Omni videos.

AIMicrosoft's MAI-Image model now ranks third on Arena's image editing leaderboard. The latest MAI-Image model is available in Private Preview on Microsoft Foundry.

AIStability AI has released two beta tools for Stable Audio 3.0: a plugin that brings audio generation into digital audio workstations (DAWs) and an upgraded experience at StableAudio.com with more editing options. Both are powered by commercially-safe models, so users own their outputs and can distribute them freely. Some features are experimental, and the company says it will keep iterating in real time.

AIStability AI released a Stable Audio plugin that runs Stable Audio 3.0 generation inside DAWs as an instrument, available as a macOS AU and VST3 with Apple Silicon and Intel support. The enhanced StableAudio.com web app adds iterative prompting, audio-to-audio variations, per-track mixing controls, and export, and both tools are in beta and powered by commercially-safe models that users can distribute freely.

AIGoogle AI Developers showcased a Chrome extension built in Antigravity with Gemini 3.7 Flash that generates rich visual definitions when users highlight text while browsing. The post says the tool uses Nano Banana and Gemini Omni to turn the web into an interactive visual encyclopedia, with a demo video linked.

AIByteDance has released Bernini-Diffusers-v2 on Hugging Face, a video generation and editing pipeline combining a Qwen2.5-VL planner with Wan2.2 diffusion components. The model card recommends it over Bernini-R for complex requests needing stronger instruction following and multi-step semantic planning. Code and weights are available under Apache License 2.0.

AIIdeogram has introduced Ad Resizer, which takes one uploaded design and exports a campaign-ready set for placements including Instagram, Facebook, YouTube, connected TV, display, and print, even at extreme aspect ratios. The tool is available now in the Ideogram app and via the API.

AIStability AI has challenged teams to build community-facing projects for musicians using Stable Audio 3.0. Teams can start now, with the code available on GitHub and model weights published on Hugging Face.

AIMiniMax launches H3, a general-purpose multimodal generation model that understands text, images, video, and audio as unified context. It generates video up to 15 seconds at 2K resolution with native stereo sound, and the company says model weights will be opened in the coming days, subject to applicable laws and regulations. MiniMax also says H3 is priced below mainstream models at 2K and 768p.

Why it matters: The post explains how a unified multimodal design and training choices enable 2K video with native stereo sound, useful for comparing against closed video generators.

AIGoogle has released short films created by its fourth Flow Sessions cohort, a six-week creative partner program for artists. The program uses Google Flow, Google's creative studio featuring its advanced generative models, to support co-creation and passion projects.

AIIdeogram has made P-Image-Ideogram available now through its API and a long list of partner platforms, including ComfyUI, Runware, Replicate, Leonardo.Ai, Picsart, Cloudflare, Magnific, Gamma, and Together AI. The company says the rollout aims to broaden access to frontier-quality image generation.

AIIdeogram highlighted four examples of its P-Image-Ideogram model demonstrating photorealism, text rendering, and style versatility. The post claims the model offers the best quality for the price and supports JSON prompting and layout control.

AIIdeogram's P-Image-Ideogram offers four quality modes (Very low, Low, Medium, High), native 1K and 2K generation, and a range of aspect ratios. Latency runs from 3s to 8s, letting users choose their point on the speed-quality curve.

AIOn Design Arena's blind human-vote image leaderboard, P-Image-Ideogram pushes the Pareto frontier on both quality versus cost and preference versus speed. The post's plain-terms reading is that any model rated higher costs more or runs slower.

AIIdeogram introduced P-Image-Ideogram, a family of image models co-developed with Pruna, offering a quality-speed-cost trade-off across four quality modes. The models generate native 1K and 2K images starting at $0.003 per image and are available now on the Ideogram API and partner platforms.

AIIdeogram's Object Remover lets users select distractions such as bystanders, power lines, or clutter and restores the scene as if they were never there. The post says nothing else in the image is altered.

AIMidjourney is releasing V8.2 today and making it the default model on the platform. The release focuses on aesthetics, personalization, and image quality, with a bolder and more creative new style. The post's two image pairs compare earlier outputs with the new style.

AIMidjourney is acquiring Co-Star, and Co-Star CEO Banu Guler is joining as Midjourney's Chief Design Officer. She will build a cross-functional design, product, and frontend team for Midjourney's efforts, while the Co-Star app continues under her full control.

AIMax Woolf says Meta's new generative image model failed immediately on his prompt injection test, which asked it to render all previous text verbatim using refrigerator magnets. He links to Meta's announcement of Muse Image. The post gives no further details about the model's capabilities or the test results.

AIMeta Superintelligence Labs has released Muse Image, which can invoke search and coding tools and self-refine its generations before output. It is available today in the Meta AI app, meta.ai, Instagram Stories in the US, and WhatsApp in limited countries, with Facebook coming soon. Meta also previewed Muse Video, which is coming soon to creators and Meta AI and is reported as ranking No. 3 on Arena for text-to-video at the time of writing.

Why it matters: The source describes how search, code execution, and self-refinement change image generation, which matters to anyone comparing agentic media models with plain prompt-to-image systems.

AITeam Motif fine-tuned Stable Audio 3.0 on Arabic maqam and built an Ableton plugin for microtonal style transfer. The project won the Stable Audio 3.0 Challenge at Music Hackspace and runs locally on device.

AIGoogle's Nano Banana account announces Nano Banana 2 Lite, which it calls the fastest banana. According to the quoted Google DeepMind post, it generates text-to-image outputs in about 4 seconds and targets quicker ideation and workflows where speed and cost matter most.

AIGoogle introduces Nano Banana 2 Lite, the fastest and most cost-efficient image model in the Nano Banana family to date. It is available now in the Gemini app and Google AI Studio.

AIMidjourney users can add --preview to a prompt for an early look at V8.2 aesthetics and personalization. The big batch draft mode now also works with --sref random, letting users explore style space 24x faster than before.

AIFei-Fei Li replied to smallfly's reflection on AI, automation, and human creativity, saying World Labs is founded on empowering human ingenuity and productivity. The post thanks smallfly for working with World Labs and links to a Fast Company article about World Labs, Marble, and spatial intelligence.

AIMidjourney has published a technical deep dive into its new "Midjourney Scanner." The post itself contains no further details on the scanner's functionality, performance, or availability.

AIMidjourney has released a big-batch draft mode for V8.1 that generates 24 lower-resolution images at half the price of a standard-resolution four-image job. Users can press "Vary" on any image they like to get new full-resolution versions.

AIByteDance has released Sa2VA-LLaVA-1.5-7B on Hugging Face, a model built on LLaVA-1.5-7B with a SAM2 grounding encoder that performs dense image and video referring segmentation alongside open-ended chat. The checkpoint is self-contained and loads with trust_remote_code=True without extra packages, and it is positioned as a LISA-comparable baseline within the Sa2VA family. Reported results include 80.3 cIoU on RefCOCO val and 54.8 J&F on MeViS (val_u).

AIByteDance's Sa2VA-Qwen3-VL-4B-SAM3 is built on Qwen3-VL-4B-Instruct with a SAM3 grounding encoder and produces dense image and video referring segmentation alongside chat. It reports 83.7 cIoU on RefCOCO val, 65.3 J&F on MeViS (val_u), and 77.1 on Ref-DAVIS17. The checkpoint is self-contained and loads on Hugging Face with trust_remote_code=True, with no extra packages required.

AIStability AI says Stable Audio 3.0 was built to support exploratory audio work such as stem remixing. A quoted post from @teropa reports using the Medium model with an init_audio input and an init_noise_level of 0.4–0.5, with empty prompts.

AIZ.ai released SCAIL-2, an open-source model that animates a reference character from a driving video without skeleton maps or inpainting masks. It also supports character replacement, multi-character scenes, and animal-driving, with 512p and 704p resolutions and inputs whose height and width are both divisible by 32.

AIByteDance has open-sourced the 1.3B-parameter weights of its Bernini Renderer (Bernini-R), available on Hugging Face as ByteDance/Bernini-R-1.3B-Diffusers. Fine-tuned from Wan2.1-1.3B, the model performs close to the 14B variant on simple tasks such as style transfer, subtitle or watermark removal, and local editing, but lags on complex tasks such as human generation. The release requires a CUDA GPU, with an H100 recommended for FlashAttention-3.

AIByteDance has open-sourced the inference code and model weights of the Bernini Renderer (Bernini-R), a DiT-based renderer paired with an MLLM-based semantic planner for video generation and editing. A diffusers-format version, ByteDance/Bernini-R-Diffusers, bundles the Wan2.2 base components with the Bernini-R transformer weights for direct loading, and the framework requires a CUDA GPU with PyTorch 2.5.1+cu124.

AIPaddleOCR and ERNIE Image are now available as official Dify plugins, bringing document parsing and image generation into Dify's agent workflows. PaddleOCR, powered by PP-OCRv5, PP-StructureV3, and PaddleOCR-VL, turns images, scanned PDFs, and multilingual documents into structured data for chunking, vectorization, and RAG, with private or on-prem deployment supported. ERNIE Image offers free generation, a Turbo mode with 8-step inference, and an OpenAI-style API.

AIFei-Fei Li praised GPIC, a new benchmark dataset for visual generation built for modern large-scale generative models. The corpus includes 100M VLM-captioned image-text pairs for training and 1M pairs for benchmarking, totaling about 28 trillion pixels. It is centrally hosted and fully permissive for research and commercial use.