A browser extension enables local image generation using SDXL models via WebGPU, running on the user's GPU without external setups. The tool supports two models: SDXL-Lighting fp16 (7 GB) and a 4-bit version (3.6 GB), with requirements including at least 8 GB VRAM for the full model and a browser with WebGPU support (Chrome/Edge 122+ or latest Firefox).
Atomistic Language Models (ALMs) unify language and atomistic structures, enabling natural language-driven crystal generation and optimization. ALMs use a continuous bridge to map language embeddings into atomistic diffusion steering space and employ Text-to-Crystal Feynman-Kac for stoichiometric accuracy. The ALM Bench benchmark evaluates text-conditioned material generation and optimization, with code and weights to be released soon.
Unlimited-OCR, a 3.3B multilingual OCR model, is available on ModelScope. It supports one-shot parsing for single images, multi-page documents, and PDFs, with full-document parsing and up to 32K output length. The model includes base and gundam image modes for diverse document layouts and supports Transformers inference with OpenAI-compatible streaming.
A study compares 2D and 3D diffusion models for generating synthetic X-ray images. It shows that 2D diffusion-based synthetic X-rays can train AI models to perform as well as models trained on real X-rays, offering a viable path to large, diverse datasets without relying on real patient data.
Boogu-Image-0.1 is an Apache-2.0 licensed open-source unified image generation and editing model family, including Base, Turbo, and Edit variants. It offers high-quality text-to-image generation, fast generation, image editing, and strong Chinese-English text rendering, with training data scale roughly one order of magnitude smaller than closed-source systems yet achieving competitive performance through improved model understanding and data quality.
NanoMaestro Realtime is a 50MB AI music model with 13M parameters that generates piano music in real time using a 2-layer LSTM. It runs locally in the browser via ONNX and Transformers.js with WASM, requiring no GPU or server backend, and works on older Raspberry Pi models.
The Moebius 0.2B image inpainting model has been successfully ported to run in the browser using WebGPU and ONNX Runtime. The project, initiated with Claude Code, converts the model's weights to ONNX and deploys them via Hugging Face, with a simple web interface available at simonw.github.io/moebius-web/.
Moebius is a 0.2B-parameter image inpainting framework that achieves performance comparable to 10B-parameter models. It is designed for lightweight, efficient image editing with minimal computational requirements.
PP-OCRv6, a new optical character recognition model, is now available on Hugging Face. It supports 50 languages and scales from 1.5 million to 34.5 million parameters, offering improved accuracy and efficiency across diverse languages.
A test evaluated 192 prompts across local text-to-image models on a GX10 Spark, assessing capabilities like text understanding, face generation, and spatial composition. Results are available on ImageBench, with comparisons to frontier APIs using vision language models, and all prompts and images are publicly accessible.
The user is seeking the current best open-source vision model that can run on an RTX 6000 Pro for OCR and classification of historical scanned documents. They note Gemma 4 31B performs well and is better than Qwen 3.6's vision encoder, asking for recommendations beyond this model.
A locally running deep neural network can turn any image into a playable game, using a small Transformer-like model trained from scratch. The model, running on an RTX 5090, generates game sequences autoregressively with real-time keyboard input, though it currently suffers from poor motion and context issues.
A pretrained speech classifier is repurposed as a backbone for guided diffusion-based speech generation. By attaching a lightweight subnetwork and training it under denoising score matching, the approach achieves high speech quality with reduced memory and computational cost, using a single model instead of two separately trained components.
FreeStyle proposes a framework that mines community LoRAs to generate large-scale style-content dual-reference image triplets. It employs a two-stage curriculum with disentanglement mechanisms to suppress style leakage and introduces a benchmark with style-invariant and VLM-based scores to evaluate content preservation and leakage rejection.
SARLO-80 is a large-scale dataset combining very-high-resolution SAR SLC, aligned optical imagery, and natural-language descriptions. It includes 119,566 triplets from 2,500 global scenes across 72 countries, standardized to an 80cm slant-range grid with pixel-level alignment and three caption variants. The dataset is publicly available on Hugging Face for multimodal learning benchmarks in native SAR geometry.
A study shows diffusion models can achieve global minimizers without explicit timestep embeddings. Ablation studies on CelebA and CIFAR-10 reveal time-agnostic models maintain high fidelity and outperform conditioned ones in FID, precision, and recall.
A pretrained speech classifier is repurposed as a backbone for guided diffusion-based speech generation. By attaching a lightweight subnetwork and training it under denoising score matching, the approach achieves high speech quality with reduced memory and computational cost, using a single model instead of two separately trained components.
A novel variable-length tokenizer uses learnable global merging to enable cross-length representation alignment in diffusion models. This data-independent approach overcomes position-dependent semantics and improves the quality-compute trade-off on ImageNet 256×25-6 generation compared to prior methods.
MakeupMirror, a diffusion-based makeup transfer model, enhances facial feature and skin tone preservation over Stable-Makeup. It achieves +60% improvement in facial recognition similarity and -50% reduction in skin tone difference, with 94% expert acceptance and 0.7s inference latency on diverse datasets.
A hybrid ANN-SNN pipeline uses pretrained EfficientNet encoders and converts their activations to spike trains via rate-coding. The system trains a CoLaNET spiking classifier with local plasticity rules, achieving 99.09% accuracy on ImageNet's 64-class benchmark, matching conventional deep networks.