AI startup Pippa pays artists royalties — but can it overcome AI ethics doubts?
AI startup Pippa pays artists royalties for using their styles, but it still relies on models trained on scraped internet content, drawing skepticism about its ethical claims.
Generative AI startups have faced criticism for training models on artists' work without consent. Now, a new company called Pippa is trying a different approach by directly compensating artists when their styles are used.
Pippa, which launched in May, offers a text-to-video tool similar to other services. Subscribers generate short video clips, but the platform's key feature is that artists are paid each time a generation draws on their style. Cofounders Hogan Shrum and Sean Wright argue that this model can prove ethical AI is possible, and that it moves past what they describe as the industry's troubled history. Wright compared the situation to the Napster era, when music was stolen until Apple introduced paid downloads.
The payments themselves are modest. Artists receive $0.005 per image and $0.003 per second of video, plus access to a 5 percent royalty pool funded by subscription revenue. Pippa's subscriptions range from $14.99 to $99.99 per month. The company compares the structure to Spotify, though many artists have criticized Spotify's royalty rates.
Currently, Pippa has around 800 paying subscribers and licensing agreements with four artists. Four more contracts may be in the works. To join, illustrators must go through a vetting process to prove ownership of their work. They can also submit under pseudonyms to avoid backlash from peers, a concern that cofounder Wright acknowledged when he noted that some artists fear being ostracized.
Despite its ethical claims, Pippa still relies on open models that have initial training on broader internet content, including scraped art. That means it cannot fully escape the very issue it aims to solve. This is a common challenge among "ethical" AI startups, as building fully proprietary datasets requires significant capital and often leads to specialized tools, as seen with Ben Affleck's InterPositive.
Looking at Pippa's generated content, there is little to distinguish it from other text-to-video services. The site mostly features children's animations that poorly imitate Pixar's style, and no output appears clearly influenced by independent artists. Pippa plans to incorporate ByteDance's Seedance 2.5 model, which can produce up to 30 seconds of fine-tunable video, but that alone won't differentiate it from competitors that offer the same model.
The founders remain optimistic, believing more artists will join as they see the payment structure. However, this depends on a broader shift in public opinion toward accepting gen AI. The company's success will likely hinge on whether it can convince artists and audiences that its methods are genuinely different.
