Google Unveils Gemini 2.0 Flash for Image Editing
New AI Models and Services
Google's Gemini 2.0 Flash Model Release: Google has released a new AI model called Gemini 2.0 Flash that can edit images, change styles, and generate new images from scratch. The model can transform images to different styles and follow complex prompts to make specific edits. However, it struggles with incorporating information from multiple images and handling human subjects accurately.
Google's Imagen 3 Model: Google has released a new AI model called Imagen 3, which is part of their Gemini AI Studio. The model is capable of editing and enhancing photos, including adding or removing hair and changing facial features. Users have reported mixed results with the model.
Gemini Model Update with Native Image Generation: Google's Gemini model has been updated to include native image generation capabilities, allowing it to create visual stories. This advancement enables the model to generate images directly from text prompts, marking a significant step forward in multimodal AI capabilities.
Gemini 2 Flash Omnimodal Image Generation: Google has released Gemini 2 Flash, the first public omni image generation model. The model is available for free and has been integrated into Google AI Studio.
AI Research and Development
Slim Attention Technique: Researchers have developed 'slim attention,' a technique that reduces the context memory size by half for transformer models with multi-head attention, speeding up inference without compromising accuracy. For encoder-decoder transformers like Whisper, the context memory can be reduced by up to 8x, significantly enhancing token generation speed.