**[Meta Launches Intelligent Agent Image Model Muse Image]** According to monitoring by Beating, Meta has launched the image generation model Muse Image, developed by its Ultra Intelligence Lab. Unlike traditional models that directly map text to pixels, it exhibits behavior patterns similar to those of a design agent. The model operates in collaboration with the recently released Muse Spark language model. Before creating an illustration, the model first formulates a composition plan. If the prompt requires factual information, it retrieves web pages to ensure accuracy; if precise typography is needed, it can render clear, readable text directly within the artwork and generate realistic QR codes based on custom prompts. Users can also use the annotation feature to mark and comment directly on the artwork, guiding the model to understand context and fine-tune local details without regenerating the entire image. Muse Image is now available on the Meta AI app and can be used for free for everyday creations, while more advanced creative needs require a paid subscription to unlock. It will also be gradually integrated into commercial systems such as Advantage+ advertising creatives. [Original Link]