Ideogram raises $80M as it releases text-to-image model


The artificial intelligence startup Ideogram announced Wednesday that it raised $80 million in new funding led by Andreessen Horowitz as it released the 1.0 version of its text-to-image generative AI model.

The Series A round also attracted investments from existing investor Index Ventures and new investors Redpoint Ventures, Pear VC and SV Angel. Andreessen Horowitz co-led the company’s $16.5 million seed funding round alongside Index Ventures in August.

Ideogram says its most recent text-to-image model is its most advanced to date and “offers state-of-the-art text rendering, unprecedented photorealism and prompt adherence.” It also comes with a new feature called “Magic Prompt,” which helps users write detailed prompts that will produce imaginative and creative images.

The new model’s claim to fame is that it is capable of producing highly reliable text rendering along with photorealistic and vivid images based on text prompts provided by users. Reliable text, or even readable text, has long been a problem for text-to-image generative AI models, which tend to produce inaccurate text, total gibberish or weird symbols when asked to produce words and sentences along with images.

To tackle these issues, major generative AI developers such as OpenAI have upgraded their models to produce more accurate results by releasing DALL-E 3, which is much more accurate. Open-source generative AI model developer Stability AI Ltd. announced a preview of Stable Diffusion 3, which also advances its text rendering accuracy.

“Our systematic evaluation shows that Ideogram 1.0 is the state-of-the-art in the accuracy of rendered text, reducing error rates by almost 2x compared to existing models,” the Ideogram team said in the announcement. The team made comparisons to DALL-E 3, showing about half the error rate in most cases.

When it comes to text-to-image generation, one thing that can really make creating images using AI for most users is prompting. This is the magic sauce behind getting a model to produce a vivid, beautiful and imaginative image. Although it’s true that models understand human speech and can be “spoken to” conversationally, the problem is that models have their own language and that the sense of what is produced must be conveyed to them in a particular way.

That’s where “Magic Prompt” comes in. According to the company, it’s a “creative assistant” that automatically “enhances, extends and translates” prompts to help users get the most out of their prompts to create beautiful and creative images.

For example, a user could put in an extremely simple prompt such as “vegetables orbiting the sun” (pictured). Magic Prompt would then take it and expand it as much as possible to get the most out of it: “A captivating and whimsical illustration of various vegetables in orbit around a bright, golden sun. The vegetables, each with their own distinct personality, are comically suspended by invisible strings, creating a sense of weightlessness. A mix of classic vegetables like carrots, potatoes, and tomatoes, as well as more exotic ones like artichokes and eggplants, fill the celestial scene. The background stars twinkle in the vast universe, adding to the fantastical, cosmic atmosphere.”

Even after Magic Prompt creates the extended prompt, the user can still modify it to fit the image that they want generated.

Image: Ideogram

Your vote of support is important to us and it helps us keep the content FREE.

One click below supports our mission to provide free, deep, and relevant content.  

Join our community on YouTube

Join the community that includes more than 15,000 #CubeAlumni experts, including CEO Andy Jassy, Dell Technologies founder and CEO Michael Dell, Intel CEO Pat Gelsinger, and many more luminaries and experts.

“TheCUBE is an important partner to the industry. You guys really are a part of our events and we really appreciate you coming and I know people appreciate the content you create as well” – Andy Jassy



Leave a Reply

Your email address will not be published. Required fields are marked *