Can Money Bridge the Artist-AI Divide?

Pippa is actively seeking to engage artists through an innovative revenue-sharing model.

For years, illustrators have vocally expressed concerns regarding generative artificial intelligence startups that train their models on artists’ work without explicit consent. They have consistently highlighted this practice as a form of theft, a claim often met by AI proponents who argue its necessity for technological advancement. This conflict has fueled numerous legal disputes but has also paved the way for a new generation of AI companies, such as Pippa, which aim to differentiate themselves by promoting more ethical services.

Similar to many platforms offering text-to-video capabilities, Pippa’s primary offering is the creation of short, user-generated video segments that can be assembled to craft visual narratives. However, Pippa distinguishes itself within the generative AI video landscape through a business model explicitly designed to compensate artists directly. Whenever a Pippa subscriber generates an image or video clip whose stylistic elements are derived from a human artist’s work, that artist receives payment from the platform.

Pippa's co-founders, Hogan Shrum and Sean Wright, are convinced that their commitment to collaborating with artists will be a key differentiator, attracting a strong customer base. They also aspire for their company to serve as proof that AI firms can operate ethically. Despite these ambitions, Pippa faces a considerable challenge in winning over the artistic community. Even as it strives to offer an ethical approach to AI video generation, the company currently relies on foundational models that are comparable to those offered by other text-to-video services.

In a recent discussion, Shrum and Wright conveyed their vision for Pippa as a means to transcend "the bloody history that the AI industry has been built on." The co-founders are determined to set a precedent for their peers, demonstrating to the public that AI companies do not inherently need to appropriate existing works to develop their products.

Wright elaborated, stating, “All of these models from the past 18 months have been trained on real people’s art without their permission — that’s how it was built.” He drew a parallel to the early days of digital music, remarking, “I harken it back to when Napster was a thing and everybody was stealing music. But then, Apple came out with the 99-cent song, Napster got shut down, and we figured out a different way to get musical artists paid.”

Individually, Pippa’s artist payments are modest, especially when contrasted with the platform’s monthly subscription fees, which range from $14.99 to $99.99. The company reports that artists receive $0.005 per image generated and $0.003 per second of video. Payments increase proportionally with user generation volume, and artists also gain access to a 5 percent royalty pool, funded by Pippa’s overall subscription revenue. Intriguingly, Pippa’s press materials draw a comparison between its royalty structure and Spotify’s — a choice that raises eyebrows given the widespread criticism from artists regarding the music streamer’s compensation practices.

Pippa’s underlying technology still incorporates art sourced from the internet.

Since its launch in May, Pippa has amassed approximately 800 paying subscribers. While the company has finalized licensing and model training agreements with only four human artists, Shrum indicated that discussions are underway to onboard an additional four. Prospective artists must undergo a multi-step vetting process to verify the authenticity and ownership of their submitted illustrations. Upon approval, models trained on their art become accessible to Pippa’s subscribers. Furthermore, the service provides artists with a profile page, enabling them to direct traffic to their work outside the platform. Wright also highlighted that Pippa offers partner artists the option to submit their work under pseudonyms, addressing concerns about being perceived as “crossing the picket line” within the artistic community.

“We’ve had so many conversations with artists where they’re like, ‘I love this, but I don’t want to get ostracized by my community,’” Wright explained. He acknowledged, “We’re very early in this space, so there’s some apprehension to participate in it. But we think that can change as people see the economics of how we’re paying artists and that our intent is good.”

Although Pippa’s long-term goal is to transition entirely to models trained on datasets provided by its in-house artists, its current technology still relies on open models that, according to the company, “have initial training on the broader set of content out there.” In simpler terms, Pippa’s technology continues to some extent to utilize art scraped from the internet without the explicit consent of creators. Despite its commendable intentions, the company has not yet found a complete solution to the inherent issue of creative appropriation prevalent in generative AI.

This challenge is not unique to Pippa; it is a critical consideration for any "ethical" AI startup whose specialized models are trained on data not generated internally. This is why some companies, such as Ben Affleck’s InterPositive, emphasize their development of wholly original, proprietary datasets. However, this approach demands substantial upfront capital, often beyond the reach of most startups, and typically results in highly specialized AI tools that are not designed for the general consumer market.

Upon navigating Pippa’s online portal, which allows users to view other people’s generated AI animations or create their own, I observed little that significantly differentiated it from other text-to-video services. Currently, the platform is largely populated with kid-oriented content that falls short in emulating Pixar’s polished aesthetic, and none of the creations truly conveyed the distinct touch of an independent artist.

This similarity is understandable given that, while Pippa intends to integrate more original art styles, it presently utilizes many of the same cloud-based models accessible through other platforms. In the near future, Pippa plans to incorporate ByteDance’s Seedance 2.5 model, which may attract subscribers due to its capability to generate up to 30 seconds of video footage that can be fine-tuned without requiring a complete restart. However, for Pippa to truly distinguish itself from competitors also offering Seedance 2.5, it will need to secure more in-house artists, a task contingent on its ability to effectively pitch its vision to potential partners.

Shrum and Wright remain optimistic about attracting more artists, believing that positive experiences with Pippa will encourage them to advocate for the platform within their respective communities. For this to materialize, however, a broader shift in public sentiment regarding the adoption of generative AI technology would be necessary.