Yesterday, I wrote about Picsart’s foray into text-to-image and text-to-text generation. I also mentioned that Picsart’s 150 million monthly active users combined with Canva’s 100 million could significantly impact how easy it is for the pure play text-to-image companies to build a large market. Thanks go to Synthedia reader Roy Murphy of Synthetic Agency for pointing out that Canva is offing eary access to a beta version of text-to-image generation today.
Canva Text-to-Image
Using text-to-image generation inside Canva is ridiculously easy. You just click the Create a Design button and then select the tab with the three dots and the More label on the left. That will pull up the text-to-image beta screen with a prompt box. Just type in your prompt and two versions of the image will be generated in the window. You can then drag the image directly over into a template to add to an existing design or begin modifying the work.
This beta feature doesn’t offer all of the controls you have access to in a Stability AI, but it is very much like the blank prompt box you find with DALL-E.
Feature or Product
The presence of pure play text-to-image companies such as OpenAI, Stability AI, and Midjourney alongside design application giants such as Canva and Picsart raises a question. Is text-to-image AI a product or a feature of a product?
This market began with text-to-image as a product. Users flocked directly to the websites of the technology developers. However, should we expect text-to-image generation to be a standalone product? Does that best serve users? It may serve them by developing deep focus and unique features for a set of users that benefit from a dedicated product. However, the larger market already has products they use in their existing design workflow. Using the image generator inside of Canva is very convenient as you already have team collaboration features and integration with your other projects and templates.
Another potential model is something you might expect from Adobe. Text-to-image could become a standalone application within its Creative Suite portfolio. With that said, a bigger benefit may be to embed the functionality as a feature inside Photoshop, Premiere, and Illustrator. It would seem to make more sense as a feature or service accessed from within these applications where they are likely to be modified after generation.
Jasper AI is treating text-to-image as a second product to complement its AI writing (text-to-text) solution. Shutter Stock is adding text-to-image as an alternative to buying an off-the-shelf stock image. What you find is that if a company has an existing customer base, they are adding text-to-image as a feature. Otherwise, they are creating a standalone product. In the end, the market will decide. Users will flock to whatever best meets their needs. That seems likely to yield scenarios where text-to-image is a product in some places and a feature in others.
Market Saturation
Text-to-image generators were a novelty just a few months ago. DALL-E, Midjourney, and Stability AI had long waiting lists to join their beta access. That meant most people had to use lesser-known solutions such as Craiyon and Night Cafe or set up their own instance of an open source model. My, how quickly the situation has changed.
Today, DALL-E and Midjourney are available for general use. More and more services such as Jasper AI are tapping into these technologies to provide a packaged solution for users. Now there is a text-to-image feature within the two largest online design applications, Picsart and Canva. The text-to-image market transformed from one of scarcity to abundance in a few months.
The next round of competition in this space will be around quality, value-added features, and niche competence. Let the games begin!