Article

Explore our curated collection of AI articles covering artificial intelligence, machine learning, AI tools and more

328 Article•Page 1 / 37

RealRestorr: Advancing Universal Real Image Inpainting with Large-Scale Image Editing Models

RealRestorr: Towards General Real-World Image Restoration with Large-Scale Image Editing Models, enabling versatile restoration of real-world images by leveraging large-scale image editing models.

Open source

MACRO: Advancing Multi-Reference Image Generation with Structured Long Context Data

MACRO advances multi-reference image generation for structured long-context data, covering four task categories—Custom, Illustration, Spatial, and Temporal—distributed across four image count brackets (1–3, 4–5, 6–7, ≥8 reference images).

Open source

Cohere Transcribe: Open-Source Speech Recognition Model

Cohere Transcribe is an open-source speech recognition model featuring a 2B-parameter architecture designed specifically for audio input and text output, supporting 14 languages.

Open source

Voxtral TTS: The Open-Source Text-to-Speech Model from Mistral

Voxtral TTS is Mistral's open-source text-to-speech model, offering fast processing, real-time adaptability, and lifelike voices for voice agents.

Open source

MinerU-Diffusion: A Diffusion-Based Document OCR Framework

MinerU-Diffusion, a diffusion-based document OCR framework, achieves up to 3.2× faster decoding speed by introducing chunk-wise diffusion and uncertainty-driven curriculum learning, while enhancing robustness and reducing reliance on language priors.

Open source

LongCat-Next, an Open-Source Native Multimodal Model

LongCat-Next, an open-source native multimodal model, integrates language, vision, and audio into a unified discrete autoregressive model, handling text, vision, and audio under a single autoregressive objective.

Open source

PrismAudio: Alibaba's Open-Source Audio Generation Model for Video

PrismAudio, Decomposed Chain-of-Thoughts and Multi-Dimensional Rewards for Video-to-Audio Generation

Open source

VideoDetective: A Plug-and-Play Inference Framework for Long Video Understanding

VideoDetective is a plug-and-play inference framework for long video understanding, which searches for clues by leveraging external queries and internal correlations.