Users have always been able to search Bing Images for an image that exists on the web, but with Image Creator they’re now able to create the images they want to see. Within Edge, users can use Image Creator in their sidebar. They’re able to create an image to share a life update with their friends online and drag and drop it within their main working page, all without losing focus on their workflow.
Today, we’d like to share a bit more detail about how Image Creator works and what steps we’re taking to ensure it remains fun and inclusive.
We’ve found that generally, Image Creator works best when you type in a description of something, with additional context like location or the art style you’d like to emulate, as opposed to a more limited description. See below how Image Creator brings “dog astronaut launching into space, digital art” to life, for example, and how that contrasts with “crochet goldfish pencil drawing."
You’re really only limited by your imagination here, so please try your own prompt, and see what the tool creates. Given that imagination is as much an art as a science, Image Creator will yield four options to choose from, that best represent what you’re looking for.
As we said last week at announce, it’s important with early technologies like Image Creator – which is powered by AI technology DALL∙E 2 by OpenAI – to acknowledge that this is new and that we expect it to continue to evolve and improve. We take our commitment to responsible AI seriously. To help prevent the delivery of inappropriate results across the Designer app and Image Creator, we are working together with our partner OpenAI, who developed DALL∙E 2, to take the necessary steps and will continue to evolve our approach. We will regularly take the feedback we have and share that with OpenAI to improve the model as well as applying to our own mitigations work.
For example, OpenAI removed explicit sexual and violent content from the dataset used to train the model. We’ve also adopted a range of mitigations, such as leveraging Microsoft Bing insights on problematic queries, as well as using blocklists and classifiers powered by Azure cognitive services that lower the risk of offensive prompts being issued. These and other mitigation efforts will continually evolve over time to ensure we're up to date with new angles of potential abuse.
Microsoft has applied additional technology to address biases sometimes found in generative image technology. While some prompts may still surface biases, this is an area we are working to continuously improve. While we’re excited to see people use Image Creator to express themselves, we also want to make sure the content created using Image Creator does not impose harm to others. This Content Policy outlines the prohibited use cases of Image Creator; if you suspect any use of Image Creator in violation of this policy, you can report it here.
We hope the steps we’re taking, and guidance above help you get the most out of the Image Creator tool and keep it fun and inclusive. We welcome your feedback and look forward to what you will create!