To see is to understand 

Video dominates today’s internet. Trillions of frames, millions of hours - and it’s growing exponentially every day. 

But understanding what’s really happening in a video – actually reasoning about it – remains one of AI’s biggest blind spots. Most systems rely on labels, tags and transcriptions. They don’t understand continuity, sequence or meaning. They don’t see video as humans do.

TwelveLabs does. Its models don’t just watch video – they read it, hear it and pick up on the rhythm of a scene. Connect a moment of dialogue to a gesture three minutes later. Track objects, movement, emotion and events through time. It’s not just vision – it’s genuine comprehension.

Based between Seoul and San Francisco, TwelveLabs is building a new kind of video infrastructure: foundation models that treat video as a living, dynamic system. As the company scaled its product offering, research teams, and global reach, it needed a new identity to express not just what the technology does, but what it means for the future of video.

Video as volume

The core conceptual leap in TwelveLabs’ technology is simple, but radical: what if video isn’t a timeline, but a volume?

Rather than processing video frame-by-frame, TwelveLabs encodes it as a multidimensional block of data – compressing audio, text, movement, visuals and context into something that can be searched, navigated and understood. It’s like giving AI not just eyes and ears, but memory, intuition and intent, too.

We turned this idea – video as volume – into the foundation for the new identity. If TwelveLabs’ models perceive time spatially, the brand should too. Threads of information pulse through layouts. Diagrams show video flowing across dimensions. Colour, movement and layout all work together to express a system in perpetual motion - smart, elegant, alive.

The language of AI video

TwelveLabs’ new identity does more than simply represent the company’s technology – it helps explain it.

At the centre is a diagram system that visualises the inner workings of the platform. Marengo is the encoder, compressing video into rich multimodal embeddings, seeing and hearing in equal measure. Pegasus is the reasoning model, traversing that data to identify events, context and intent. These models go beyond labelling – they interpret the ‘what’ and get to the ‘why’. 

Pentagram built visual tools that surface this complexity – thread-based diagrams that scale from frame to library, from specific clips to sweeping concepts. They appear throughout the product and communications, flattening the learning curve and turning deep AI infrastructure into something visible, graspable and legible.

The interface itself participates in this visual logic. Search results appear as threads – modular video blocks arranged in time and space. Users don’t just see answers, they see meaningful relationships.

It’s a system that visualises intelligence, and helps others understand it.

Motion with meaning

TwelveLabs’ symbol, a galloping horse, nods to Muybridge’s early motion studies, but also signals the forward energy of the company itself. A marker of movement and intelligence, not just speed.

We reinterpreted the horse as a modular, dimensional system that could animate through layouts, inform model architecture names and give rhythm to the wider brand.

Motion across the brand behaves like the platform itself: fluid, layered and continuous. No idle loops. No decorative transitions. Everything moves with intent - revealing structure, reinforcing understanding and flowing seamlessly between surfaces.

A thoughtful game-changer

The verbal identity reflects TwelveLabs’ grounded, purpose-led approach. This is not a brand trying to be loud or disruptive for the sake of it. It’s a brand building breakthrough foundational technology with real focus, ambition and care. 

The tone of voice is confident without being inflated, fluent in the technology, without being too obscure or alienating. That’s because TwelveLabs needs to speak to developers and researchers, but also to enterprise partners, creative teams and curious users. So we created the voice for a brand that knows what it’s doing, quietly reshaping how we interact with the moving image. 

The same philosophy shaped the identity system: a layout grid based on threads and timelines. A lozenge-shaped interface that mirrors the structure of video data. A typeface that balances mechanical structure with human softness. A colour system grounded in logic but expressive in use.

An infrastructure brand, with soul

Most AI brands fall into two camps: cold and corporate, or chaotic and playful. TwelveLabs needed something else. A system with depth, precision and presence – but also emotion, story and a strong point of view.

We built a design language that feels foundational and expressive. System-led but flexible. Built to scale, but never generic. It reflects the company’s core offering: intelligence with context, clarity with feeling, movement with meaning.

TwelveLabs helps machines understand video like people do. This identity helps people understand TwelveLabs.