A little over a month after releasing its advanced text-to-image model, Ideogram has launched an update that introduces several new features, including description-based referencing and negative prompting.
These enhancements, available on Ideogram’s web platform, aim to provide users with greater control over image creation while improving the quality and coherence of the outputs. This update represents a significant step toward competing with established rivals in the image generation field, such as Midjourney and DALL-E.
What's New in Ideogram?
With the initial launch of version 1.0 in February, users gained access to a magic prompt feature that enriched their input. Building on this foundation, Ideogram has now introduced a new Describe capability that generates captions from reference images.
Users can upload an Ideogram-generated public image or their own, prompting the AI to produce a text-based description. This description can then be refined to create a similar image tailored to specific needs.
Additionally, Ideogram is rolling out negative prompting, allowing users to indicate what they do not want in their outputs. This feature helps users eliminate certain objects or styles from the final generation.
Furthermore, users can choose among Fast, Default, or Quality modes for output generation. The Fast mode produces basic images in about five seconds, while the Quality mode focuses on photorealism over roughly twenty seconds. The Default mode strikes a balance, generating images in about twelve seconds.
While user adoption of these modes is yet to be seen, Ideogram encourages use of these options to generate a basic image quickly and then refine it for higher quality results.
Enhanced Photorealism and Text Rendering
Ideogram is also improving text rendering capabilities, boasting a 15% reduction in error rates. While this change may seem modest, the company claims that it outperforms DALL-3 Vivid in generating characters and words.
Though no statistics comparing the updated model with Midjourney have been shared, Ideogram asserts that the latest version offers enhanced image coherence and photorealism, with human raters preferring it 30-50% more than its predecessor in prompt alignment, image coherence, and text rendering quality. Since launching public beta last year, Ideogram has attracted over seven million creators.
Currently, negative prompting and speed modes are exclusive to users on Ideogram's Basic and Plus plans. The availability of the reference image captioning feature remains unclear, though it may be free, akin to the Remix feature. The enhancements in text and image coherence are accessible to all users.
Ranking
