Ion Video restructures video to enable AI to supercharge its use

Australian tech company Ion Video says it has solved one of the biggest structural limitations of the internet: video’s incompatibility with Artificial Intelligence (AI).

The issue for AI is that it can analyse a video and identifying objects, scenes or timestamps but in a major limitation it cannot easily assemble new video experiences from existing content. Any modification requires editing and re-rendering, which creates new files and drives significant storage, compute and workflow costs.

ASX listed Ion Video’s approach centres on treating video as a single, fixed file, with its patented technology separating the internal structure of the video from the raw audio and visual data that make up the content.

By separating these components, Ion can create a small virtual representation of the video’s structure while leaving the original media data untouched.

This virtual file acts as a blueprint that tells AI systems how to reconstruct the video dynamically. Instead of generating a new rendered file each time content is modified, AI can assemble the required sequence of frames on demand.

The result is that video becomes programmable.

The technology was patented several years ago by Melbourne-based Australian innovator Finbar O’Hanlon (pictured) but has only come to the fore in recent years with the advent of AI.

In 2009 O’Hanlon filed a series of patents that laid the foundations for this approach, leading to the creation of Linius Technologies, which subsequently listed on the ASX.

Finbar O’Hanlon is now leading innovation at Ion Video which offers a re-engineered and expanded version of that earlier concept, incorporating new patents and adapting the technology to the rapidly evolving AI landscape.

O’Hanlon said the issue is that despite enormous advances in streaming, cloud computing and AI, the fundamental structure of video has barely changed in decades.

“Traditional video files are designed as completed, rendered assets. Once produced, they become static objects intended purely for playback and distribution. Platforms can compress, stream and analyse them, but they cannot easily manipulate or recombine their internal components without creating entirely new files.

“That design made sense in the broadcast era, when video was created once and distributed to mass audiences but AI operates very differently.

Machine learning models developed by companies such as Google, Microsoft, OpenAI, Apple and Meta Platforms increasingly rely on flexible data structures that can be recomposed dynamically.

O’Hanlon said text, images and code can already be manipulated in this way. Video cannot.

According to O’Hanlon, this architectural mismatch is becoming a major bottleneck as AI systems evolve. “Once video is rendered, it becomes a sealed object,” O’Hanlon said. “It was designed for playback and distribution, not for intelligent systems to reorganise, recombine or compose with.”

He said that when the video becomes virtualised, AI systems can interact with it in fundamentally new ways.

“Instead of editing or generating new files, an AI model can dynamically assemble content based on instructions or prompts.”

For example, a user could request: “Show me five Asian recipes under $15.”

AI could scan multiple cooking videos, identify relevant scenes and assemble a customised video sequence containing only the necessary steps.

A second instruction might refine the output further, removing commentary and leaving only the cooking process and finished dishes.

Because the video is constructed dynamically using the virtual structure, the system does not create a new rendered file. Instead, it simply references the existing video data and arranges it according to the blueprint.

For AI developers, that effectively turns video into something more like language, a medium that can be recomposed dynamically rather than consumed as a static asset.

“For the first time, AI can work with video the way it works with text,” O’Hanlon said.

What is attractive to Ion Video is taking into account the scale of global video content which makes up 82% of all internet traffic. Platforms such as YouTube receive roughly 720,000 hours of new video uploads every day, creating one of the largest repositories of human knowledge and culture ever assembled.

With video archives virtualised and made programmable, AI can assemble entirely new video experiences on demand.

Rather than browsing through hours of footage, users could simply describe what they want to see and have the system construct a tailored video sequence from existing content.

“This shift could fundamentally change how video is consumed across education, media, training and enterprise environments,” said O’Hanlon.

He said it was important to point out that Ion Video is not positioning itself as a consumer-facing platform.

“Our IP and technology is an infrastructure layer that sits beneath existing video ecosystems. The aim is to help enable hyperscale cloud providers, AI developers and streaming platforms to innovate and integrate intelligent programmable video capabilities.

“In this model, Ion’s technology acts as a bridge between AI systems and video libraries.”

The company believes the economic benefits could be substantial.

Ion Video’s aim is to licence the hyperscalers for its use, with a commercial model based on enablement value. “We believe we can save upwards of 70% in transcoding, storage and compute costs for processing video,” said O’Hanlon.

“That matters when you look at the scale of infrastructure spend. Alphabet’s capital expenditure is projected at $175 to $185 billion in 2026, driven almost entirely by AI infrastructure demand and video represents over 80% of all internet traffic. If we can save even 10% of the video-related portion of that spend, and we charge between 3 and 5% of that enablement, the numbers speak for themselves.”

The company has also filed additional patents extending its original intellectual property around video virtualisation, aiming to align the technology with the emerging wave of AI-driven media experiences.

According to O’Hanlon, the timing of the technology is critical. As AI evolves into systems that understand human intent and shape digital experiences around individuals, the ability to dynamically assemble video could become an essential capability.

If Ion’s approach proves viable at scale, it could mark a fundamental shift in how video is created, distributed and consumed.

Instead of static files designed for playback, video could become a flexible data layer that AI can assemble and personalise in real time thus redefining the next phase of the internet’s most dominant medium.