Skip to content
K = 0.731
All news · Today1 items

#Image generation

All collected and published items, ordered by time. Picks are a selected subset.
9/30Wed
  1. Google Research42

    Google Research introduces Diffusion Controller to unify and simplify AI image-generation control

    Google Research introduced Diffusion Controller, a framework that recasts the denoising process in diffusion models as a continuous control problem. A lightweight “steering damper” network dynamically adjusts the generation trajectory while keeping the base model frozen. On Stable Diffusion v1.4, its fully unlocked version achieved a 90% win rate against the baseline on the HPS-v2 evaluation. It also supports custom control of closed-source models whose internal weights are unavailable.

9/24Thu
9/23Wed
  1. Qwen · X62

    Qwen-Image-2.1 ranks first among open models on Image Edit and Text-to-Image Arena

    Alibaba's Qwen team says Qwen-Image-2.1 has become the top-ranked open model on both the Image Edit and Text-to-Image arenas. Arena data gives it a score of 1,367 on Image Edit Arena, first among open models and 16th overall, three points behind GPT-Image-1.5-high-fidelity in 15th place. The team invites users to try it and share their results.

9/22Tue
9/21Mon
  1. Qwen · X58

    Qwen-Image-2.1 gets day-one support in SGLang-Diffusion

    Qwen says SGLang-Diffusion now supports Qwen-Image-2.1 for text-to-image generation, multi-image editing, and transparent RGBA output. SGLang says native-precision inference with CPU offload runs on a single 24 GB RTX 4090, taking 18.7 seconds for 1024×1024 generation and 21.7 seconds for image editing, with peak request-time memory of 22.7 GiB. On an RTX PRO 6000 96 GB, generation and editing take 8.0 and 9.6 seconds. Inference supports TP/SP, LoRA, and an OpenAI-compatible API, with 40 denoising steps, one image per request, and no quantization.

9/17Thu
  1. Gemini App · X42

    Go from inspiration to physical objects with Gemini. With Canvas in Gemini, you can code a custom app to design a vase and change its mathematical parameters to visualize it in 3D. Then, export it as an .STL file for 3D printing to watch your ideas take shape.

    Go from inspiration to physical objects with Gemini. With Canvas in Gemini, you can code a custom app to design a vase and change its mathematical parameters to visualize it in 3D. Then, export it as an .STL file for 3D printing to watch your ideas take shape.

9/11Fri
  1. Google DeepMind · X18

    Google DeepMind uses AI to recreate a memory that was never filmed

    How do you recreate a memory that was never filmed? Our team combined restored archival photos with pose-controlled models to capture Burt and Ethelle's mannerisms and micro-expressions. That helped recreate the day they first met for *Love, Rendered*, a new documentary made with @PrimordialSoup_ and @StorySyndicate_. Watch the full film on YouTube → https://goo.gle/3V8m94m

9/10Thu
9/8Tue
  1. OpenAI News60

    OpenAI releases ChatGPT Images 2.5

    OpenAI has released ChatGPT Images 2.5, which turns ideas, sketches, and reference photos into more personalized, polished images that better match what users had in mind.

    Why it matters: OpenAI's official launch of ChatGPT Images 2.5 gives readers an update on its positioning and input options for personalized image generation.

7/23Thu
  1. Hugging Face Blog62

    Diffusers adds native support for Nunchaku 4-bit diffusion inference

    Hugging Face has added Nunchaku Lite to Diffusers, allowing users to load Nunchaku SVDQuant 4-bit checkpoints directly with `from_pretrained()` without a custom pipeline or local CUDA compilation.

    Why it matters: The article reports how to load Nunchaku Lite in Diffusers, along with VRAM and latency comparisons, helping readers assess the practical requirements for running diffusion models on consumer GPUs.

7/16Thu
7/1Wed
5/3Sun
4/23Thu
  1. Google Research60

    Google Photos adds Auto frame, using generative AI to change the viewpoint within a photo

    Google Research has announced a way to adjust viewpoint after a photo is taken in Google Photos' Auto frame feature. It treats the photo as a 3D scene and automatically changes the virtual camera position and focal length. The two-step method first uses an internal 3D point-map estimation model to reconstruct people and faces and estimate the original camera focal length, then uses a generative latent-diffusion model to fill blank areas revealed by the new view. For photos with people, the feature now automatically offers a second, reconstructed version that can correct perspective distortion in wide-angle selfies.

    Why it matters: Google Photos combines 3D-scene estimation with generative inpainting, showing how a new post-capture viewpoint-editing workflow is implemented.

1/10Sat
  1. Berkeley AI Research24

    Berkeley AI Research proposes an information-driven framework for imaging-system design in NeurIPS 2025 paper

    A Berkeley AI Research team proposes a mutual-information-based framework for evaluating and optimizing imaging systems. It estimates information directly from noisy measurements to predict system performance. Validated in color photography, radio astronomy, lensless imaging, and microscopy, optimized designs reach end-to-end-method performance with lower memory and compute needs and no task-specific decoder. The paper appeared at NeurIPS 2025.