Recently, the image AI startup Recraft released its latest text-generating image model, Recraft v3, which has shown exceptional performance in independent testing, redefining the standards for image generation performance. According to Recraft, the v3 model excels in generating text within images, maintaining anatomical accuracy, understanding prompts, and creating high-quality visual content. Its biggest breakthrough is the ability to accurately present long segments of text in a single generation, while many other models struggle with more than a few words.

Recently, a research team jointly launched an open-source AI image generation model named Meissonic. Surprisingly, this model uses only one billion parameters while being able to generate high-quality images. This compact design gives Meissonic the potential to enable localized text-to-image applications on mobile devices. The technology behind this model involves a research team that includes Alibaba, Skywork AI, and researchers from various universities. They employed a unique transformer architecture and innovative training methods, which allow

Midjourney recently announced a series of exciting updates and upcoming new features, indicating the company's rapid development and innovation in the field of AI image generation. Firstly, Midjourney is set to launch a new online editor that will allow users to import images and edit them using built-in depth maps and image restoration features. The team is actively discussing what features will be included in the initial release and is optimizing the user interface to differentiate the new features from the existing image functionalities. Additionally, content moderation...

The image generation field has welcomed a striking new star. A mysterious model named "Blueberry" has suddenly appeared in the AI arena, rapidly rising to the top with its astonishing performance, sparking widespread attention and discussion in the industry. This model, called "Blueberry," has outperformed well-known competitors, including OpenAI's "Strawberry," Flux.1, Ideogram v2, and Midjourney v6.1, swiftly becoming the new champion in the image generation domain. However, the true identity of "Blueberry" remains unclear.

David Holz, founder and CEO of Midjourney, recently shared the latest progress of the company's current projects during 'Office Hours' on Discord and addressed community user questions. This communication revealed that Midjourney is working on multiple technological innovations to compete in the increasingly intense field of AI image generation. Although the new version has been delayed, it promises richer features. Despite rapid progress in a short time, the frequency of new model releases has noticeably slowed down recently.

Freepik's AI image generation model Mystic, in collaboration with Magnific AI, has ushered in a new era of AI creation. This model not only generates ultra-realistic images but also captures intricate details in complex scenes, even generating realistic text within images, achieving full HD quality. The uniqueness of Mystic lies in its integration of the strengths of Stable Diffusion, Flux, and Magnific's proprietary models, finely tuned by a professional team.

Getty Images has partnered with NVIDIA to launch an upgraded AI image generation tool, which boosts the speed to generate 4 images in just 6 seconds, doubling the speed of the previous model. The new model employs the NVIDIA Edify architecture, significantly improving output quality, supporting longer instruction inputs, enhancing image upscaling, and allowing model fine-tuning. The tool introduces an AI image editing feature, enabling users to modify generated AI images and pre-shot creative images, including the addition or alteration of elements and canvas expansion. This feature is now live on iStock and will soon be available on other platforms.

In a pivotal copyright infringement lawsuit, a judge ruled that visual artists' lawsuits against well-known AI image and video generation companies can proceed and enter the discovery phase. The artists accuse these companies of using their works without permission to train AI models, including Midjourney, Runway, Stability AI, and DeviantArt. The crux of the case is whether the AI companies constitute 'induced infringement.' The judge indicated that the allegations are strong enough to warrant further investigation.

iFlytek announced a significant upgrade to version 2.0 of its Smart Language tool, bringing revolutionary changes to PPT creation. This intelligent tool, based on the native iFlytek Spark AI application, dramatically enhances the efficiency of presentation creation with its one-click PPT generation feature. The new version has achieved substantial upgrades in three technological aspects: PPT text generation, AI PPT layout creation engine, and AI PPT online editing module, allowing users to create PPTs in a more free, personalized, and professional manner.

FLUX.1, an open-source large model platform launched by AI image generation expert Robin Rombach, features a Vision Transformer architecture with 12 billion parameters. Utilizing techniques such as flow-matching training methods and rotational positional embedding, it significantly surpasses closed-source models like DALL·E 3 and Midjourney V6, as well as the SD3 series. In terms of performance, FLUX.1 excels in text semantic restoration, image quality, action consistency, coherence, and diversity.

The leading domestic AI image generation platform "LiblibAI" has completed three rounds of financing in the past year, totaling hundreds of millions of RMB, setting a record for the largest total financing in the domestic AI image sector. These three rounds include an angel round led by Source Code Capital, a second round led by strategic investors, and a third round led by Mingshi Capital, with existing shareholders continuing to increase their investment. Yuan Shi Capital served as the exclusive financial advisor in multiple financing rounds.

Researchers from Stanford University and the University of California, Berkeley conducted an experiment to explore the consequences of retraining AI image generation models using images generated by the AI itself. The experiment utilized the Stable Diffusion (SD) model and selected high-quality facial images from the FFHQ dataset for automatic generation. The results showed that after retraining, the AI model produced highly distorted images. This distortion originated not only from the text prompts used for retraining but also persisted even when retraining was done solely with real images.

In January 2024, Leonardo.ai surpassed Midjourney to become the number one AI image generation product. In February, Midjourney experienced a decline in traffic. Leonardo.ai demonstrated stronger performance, while Midjourney faced pressure in maintaining traffic. Midjourney has issues with its image generation model and user retention, and its usage costs are relatively high. Leonardo.ai excels in user engagement and session duration, with a steady increase in unique visitors.

CEO Sundar Pichai criticized the image generation errors of Google Gemini AI, calling them 'completely unacceptable.' The team is correcting the issues and emphasizes the importance of providing fair and accurate information, with structural reforms underway. The article was published on 2024-02-28.

Users criticized Gemini for generating diverse images in inappropriate historical contexts. Google urgently fixes the AI tool to avoid accusations of racial discrimination, as the Gemini bot had previously faced diversity issues.

iStock's Generative AI: Getty Images and Nvidia collaborate to launch Generative AI by iStock. Designed for small and medium-sized enterprises, this platform aims to help SMEs more efficiently acquire the photos they need, offering plans for individual or single-user usage. Pricing and legal protection: the platform offers a service of 100 generations for $14.99, generating four images per request.

The Alibaba Tongyi Laboratory open sources the AI image generation fine-tuning framework SCEdit. The SCEdit framework supports fine-tuning for image generation tasks and demonstrates high efficiency in generation tasks. It achieves precise control in controllable generation tasks while saving training memory costs.