Shutterstock Integrates With DALL-E for AI-rendered Stock Photos
The rising tide of AI-rendered images keeps making headway despite some efforts to stymie it. The latest gain for AI-generated photos is a recent announcement by stock photo giant Shutterstock that it’s collaborating with the creators of the DALL-E AI platform.
The collaboration will basically let users generate synthetic images with the OpenAI DALL-E platform through Shutterstock by typing in keyword prompts and then downloading the results directly.
The new DALL-E/Shutterstock image rendering feature is expected to launch in the next few months.
Without a doubt, this move will feel like a slap to the face for thousands of stock photo creators who have submitted their real, human-created images as stock archives to the Shutterstock site.
However, this is how technology sometimes progresses, and it’s yet to be seen how well the feature takes off or how much it displaces sales of real stock photos.
For those of you who haven’t fully entered the look about what AI image rendering programs are and how they may affect photography, we’ve covered that here.
The bottom line is that programmers teach machine-learning algorithms to generate images from text prompts by first feeding them millions of real-life photos and their descriptions until the algorithms (the AI part of the process) learn how to render their own based on what they’ve been previously shown.
Thus, if you prompt one of these platforms (and there are many of them beyond just DALL-E) to create a photo of a “photographer taking a photo of a model”, that’s what it will try to do.
As our own experiments in exactly this have shown though, the results created by the AIs aren’t always exactly stellar.
Either way, this will be the option available to users of Shutterstock who want quick stock images for their own needs. Even if the results can sometimes look laughably strange, the quality of the images generated by these AI platforms is steadily improving. Carefully phrasing text prompts can also create decently usable results.
Shutterstock, being “mindful” of the fact that thousands of photographers have unknowingly had their real-life photographic work used to teach DALL-E, has created a fund to compensate said photo owners.
As the site explained in a press release, “Given the collective nature of generative content, we developed a revenue share compensation model where contributors whose content was involved in training the model will receive a share of the earnings from datasets and downloads of all AI-generated content produced on our platform,”
How exactly this compensation will be broken down and how much the amounts awarded to individual photographers will amount to isn’t clarified yet.
Also, because the copyright laws surrounding AI-generated images created by algorithms trained with real photos are still fuzzy at best, it seems that Shutterstock’s revenue-sharing share model is more of a PR hedge than a real basis for notable photographer earnings.
The company further elaborates by saying, “Shutterstock has also created the framework to provide additional compensation for artists whose works have contributed to developing the AI models. The company also aims to compensate its contributors in the form of royalties when their intellectual property is used,”
OpenAI’s CEO Sam Altman has also chimed in by clarifying that “The data we licensed from Shutterstock was critical to the training of DALL-E,” He also added that “We’re excited for Shutterstock to offer DALL-E images to its customers as one of the first deployments through our API, and we look forward to future collaborations as artificial intelligence becomes an integral part of artists’ creative workflows.”
Shutterstock also said that it will aim its R&D on creating and publishing insights into how AI-generated art is created by these algorithms.
Despite all the above, it still remains unclear how the company plans to determine what photos by which photographers had what levels of influence on AI-rendered photos in a way that can be used to define compensation specifics.
Many of the thousands of photographers who have submitted their real photos to Shutterstock over the years might just find that last detail important.
Previously, both Getty Images and Shutterstock banned AI-generated images from their platform to one degree or another. This more recent move by Shutterstock seems like an attempt to more directly profit from the inevitable proliferation of these images and the technology behind them,.
It’s also worth noting that while users will be able to create their own AI-rendered visuals on the platform, uploading AI-rendered images from other sources to Shutterstock is still forbidden.
Doing this is obviously aimed in part at protecting the company’s own AI sales strategy, but it will supposedly also, let Shutterstock identify whose real photos were used for an AI image and pay those photographers from its compensation fund. Payments to photographers whose work was used by the DALL-E AI will be paid every 6 months for the use of their images in both training data and from image royalties.
The company hasn’t yet elaborated on how it plans on charging users who want to create and download images.