Pippa: Can Ethical AI Win Over Skeptical Artists?

Pippa: Can Ethical AI Win Over Skeptical Artists?

Illustrators have spent years sounding the alarm about generative artificial intelligence startups that train their models on artists' works without permission. They have pointed out that this practice amounts to theft, and in response, many AI advocates have argued that it is necessary for the evolution of technology. This has led to controversial legal battles but has also given rise to a new wave of AI startups like Pippa, which position themselves as more ethical than their competitors.

Like most companies offering text-to-video models, Pippa's main product consists of short sequences of user-generated videos that can be assembled to tell visual stories. However, what sets Pippa apart from other players in the generative video sector is how its business model is designed to directly compensate artists. Every time a Pippa subscriber generates an image or video clip whose style is derived from a human artist's work, that artist is compensated by the platform.

Pippa's co-founders, Hogan Shrum and Sean Wright, believe that their focus on artists is what will attract customers. They also believe that their company will prove that AI companies can generate content ethically. However, Pippa faces a significant challenge in convincing artists. Even while trying to create an ethical angle on AI-generated videos, the company still uses the same poorly curated models offered by other text-to-video services.

In a recent conversation with Shrum and Wright, they stated that they see Pippa as a way to move beyond "the bloody history on which the AI industry has been built." The duo aims to set an example for their peers and show the public that AI companies do not necessarily have to steal to make their products work.

Pippa does not compensate artists significantly on an individual basis—especially compared to the fees the service charges users for monthly subscriptions (which range from $14.99 to $99.99). The company states that artists receive $0.005 per image and $0.003 per second of video. The more content a user generates, the more an artist's payments increase, and Pippa also gives them access to a share of a 5% royalty pool funded by the platform's overall subscription revenue. In its press materials, Pippa compares its royalty payment structure to that of Spotify—a curious choice given the number of artists who have complained about being shortchanged by the music streaming service.

Launched in May, Pippa currently has around 800 paying subscribers, and although the company has only signed licensing/model training agreements with four human artists, Shrum stated they are in talks to bring on four more. Artists interested in working with Pippa must first go through a multi-step verification process to ensure that the illustrations they submit genuinely belong to them. Once an illustrator is approved, the models trained on their art become available to Pippa subscribers. Additionally, the service provides artists with a profile page, which can be used to direct people to their work outside the platform. Wright noted that Pippa also allows its partner artists to submit their work under pseudonyms if they fear being perceived as "betraying their community."

While Pippa hopes to transition entirely to models trained on datasets provided by its in-house artists, this technology still relies on open models that the company claims "were initially trained on a broader content set." In other words, Pippa's technology still involves some use of art that has been scraped from the internet without the explicit consent of the creators. The company's intentions may be good, but it does not really have a solution to fully address the inherent creative theft associated with generative AI.

This issue is not unique to Pippa; it must be considered by every "ethical" AI startup whose custom models are trained on information that has not been created in-house. This is why some companies—like Ben Affleck's InterPositive—have emphasized that they are developing entirely original and proprietary datasets. However, this requires a lot of initial capital that most startups do not have access to, and the end result is usually highly specialized AI tools that are not exactly designed for general consumers.

While browsing Pippa's online portal where you can view animations generated by others or create your own, I did not see much that seemed very different from what one might find on other text-to-video services. Currently, the site is primarily filled with content aimed at children that poorly mimics the polished aesthetic of Pixar, and nothing really resembles what would have been inspired by the hand of an independent artist.

This makes sense given that, while Pippa intends to incorporate more original artistic styles, it still uses many cloud-based models that other platforms offer. In the near future, Pippa plans to incorporate ByteDance's Seedance 2.5 model into its service, which could attract some subscribers due to the model's ability to generate up to 30 seconds of video sequences that can be adjusted without starting from scratch. But for Pippa to truly stand out from competitors that also offer Seedance 2.5, it will need to sign more in-house artists, which will depend on how the company is able to present itself to potential partners.

Shrum and Wright believe they will find more artists to work with, and that using Pippa will convince these artists to sell the idea of joining the platform to their communities. For this to happen, however, public opinion on the use of generative AI will need to evolve in favor of the technology.