It’s a problem as old as time: How do you train your text-to-image AI to create coherent text synthesis? Well, okay maybe that’s a relatively new question. But even with AI still being in its earliest stages, it has been an issue that has kept other AI apps from tackling too much text rendering.
However, as VentureBeat reports, there’s a new AI image startup in town called Ideogram which claims to have solved these issues with text generation and rendering. Let’s look at Ideogram, how it compares to other text-to-image AI apps like Dall-E 2 and Stable Diffusion, and ultimately explore how its text rendering could be used for video-centric text generation and animation.
Introducing Ideogram AI
Founded by former Google Brain researchers and recently launched with $16.5 million in seed funding led by AI-focused firms like a16z and Index Ventures, Ideogram is here to shake up the text-to-image AI space thanks to their breakthrough in text generation.
Operating currently as a web app, which you can find at ideogram.ai, the company appears to be another essential play in this AI image generation space with results seeming quite similar to MidJourney, Dall-E, Stable Diffusion, etc…
However, as the company lays out on its site, Ideogram’s killer feature is its solution to the problem of creating reliable text generation within images. This issue has plagued other AI apps for some time now.
AI Powered Text Generation
So, how does the text generation in this new Ideogram AI work? When you log into the web app you can browse all types of examples and explore different prompts, but the one to check out is the one labeled “typography,” which can render lettering in any variety of different colors, fonts, sizes, and stylings.
You can of course still try out other preset styles like painting, fashion, product, illustration, and even 3D rendering and cinematic options, however, this typography prompt really unlocks everything an AI creator would need for developing lettering for signs, ads, and company logos.
A test of the Ideogram's AI text generation
The Future of AI Text Generation
Judging from early reports though, this text generation doesn’t appear to be perfect just yet. (As you can see in the example above and the cover image for this article, the web app currently struggles with spelling — even when you provide the text for it in the prompt.)
Ideogram is only currently launched as a beta right now after all. However, the promise of what’s to come for this AI tech, and AI in general, in this text generation space should be intriguing to any video editors or motion graphic artists.
Text in video has a long history that goes far beyond subtitles for foreign films. Anyone who’s spent hours rendering text or logo animation in After Effects can tell you, that if there were AI solutions for streamlining this text generation and animation process, they would be welcome additions to any post-production workflow.
From the looks of it, Ideogram is focused on single-image solutions right now. However, as we’ve seen with other AI apps launched in the past few months, it’s not a long runway (yup, even them, too) for text-to-image to become text-to-video.- How To Replace Your Actors With a CG Character With Just One Click ›
- Even More Generative AI Video is on Its Way with Stable Video Diffusion ›
