STOCKHOLM, SWEDEN, 01 October 2026.
Vionlabs, the AI company behind the entertainment industry’s most advanced content understanding models, today announced Curio by Vionlabs, a new personalization layer built directly on top of Fingerprint Plus, its proprietary multimodal content understanding platform. Pure Flix is the first platform partner to bring Curio to its members.
Personalization, Not Just Another Recommendation Engine
Weak discovery has a real cost: streaming viewers spend an average of 14 minutes searching for something to watch, 19% abandon the session when that search comes up empty, and nearly half say they're willing to cancel a service over difficulty finding something to watch[1]. That’s rarely an algorithm problem. It's a foundation problem. Most recommendation systems are built on sparse inputs – genre, cast, release year, a handful of editorial tags – and can only ever learn categorical similarity. Two thrillers look identical even when one is a slow-burn character study and the other is nonstop action.
Curio starts from a different foundation entirely. Rather than mapping content down to simplified tags, it runs natively on Fingerprint Plus embeddings, Vionlabs’ proprietary representation of a title’s visual composition, pacing, dialogue, tone, and emotional arc, learned from more than 1,000,000 analyzed assets. That means Curio isn’t matching viewers to categories, it’s matching them to how a title actually feels to watch. This is what makes it possible to personalize accurately even for new and long-tail titles with no viewing history yet.
Because Curio is an extension of the Fingerprint Plus foundation customers already run, not a separate system layered on top, it can be deployed without rebuilding a content understanding layer from scratch, and existing customers can activate it as a natural next step rather than a new integration.
“This isn’t ‘recommendation’ as usual. It’s the next expression of what Fingerprint Plus already understands about content. We built the deepest content understanding layer in the industry, and personalization was always the natural next step. Now that same understanding carries straight through to what a viewer actually sees on their screen.”
Arash Pendari, Founder & CEO, Vionlabs
How It Works
Under the hood, Curio uses a two-tower architecture, the same modeling approach used at scale by platforms like TikTok, YouTube, Meta, and Instagram to power personalized feeds and discovery. One tower learns a representation of the viewer; the other learns a representation of the content; Curio matches the two within a shared embedding space.
The architecture itself is well understood in consumer tech. What sets Curio apart is what feeds both towers: instead of sparse tags, it runs directly on raw Fingerprint Plus embeddings, interpreting content and viewer behavior at the semantic level with no lossy mapping down to simplified categories in between. The result is a system that connects titles by resonance (pacing, tone, emotional arc) rather than by category alone, and performs meaningfully better on cold-start and long-tail discovery than collaborative-filtering-only approaches.
Curio is also privacy-first by design: it operates on pseudonymized viewer IDs only, so no personal identifiers ever need to leave a customer’s own environment, with integration built to maintain full GDPR compliance throughout. Customers can start with a lightweight Watch History API for fast, stateless integration, or move to a full deployment that builds long-term behavioral profiles for deeper personalization, along with expanded recommendation types and performance analytics.
Pure Flix: From Content Understanding to Personalized Discovery
Vionlabs has worked with Pure Flix for several years, powering the platform’s editorial curation through Fingerprint Plus. With that foundation in place, Pure Flix is now activating Curio across three core experiences: Personalized Recommendations, Because You Watched, and Similar Titles on title detail pages.
Availability
Curio launched at IBC 2026 in Amsterdam (September 11–14) at Vionlabs’ booth in Hall 14, Future Tech, and is available now as an add-on for existing Fingerprint Plus customers, and as part of a full-stack deployment for new partners.
About Vionlabs
Vionlabs builds proprietary AI models trained specifically on entertainment content, helping streaming and media platforms understand, curate, and personalize their content at scale. Vionlabs’ models have analyzed more than 1,000,000 assets, powering editorial curation, thumbnail and clip generation, content operations, and, through Curio, personalization for leading OTT platforms worldwide.
Media Contact
Sources
[1] Gracenote, 2025 State of Play, November 2025. Based on a survey of 3,000 consumers in Brazil, France, Germany, Mexico, the U.K. and the U.S.
.png)
