For years, digital video advertising has followed a relatively simple model: a video plays and, at certain moments, another advertising video is inserted.

In linear television this process was fully integrated into the broadcast signal. In streaming, however, the arrival of HTTP-based architectures, segmentation, and CDN-based distribution has forced the industry to reinvent the way ads are inserted.

In previous articles we have already analyzed the dominant models: Client-Side Ad Insertion (CSAI), Server-Side Ad Insertion (SSAI), and even its evolution, Server-Side Guided Ad Insertion (SGAI). At the same time, new debates are emerging within the industry around a possible conceptual evolution:

What if advertising stopped being a video inserted into another video and instead became a function executed within the streaming data flow itself?

Although this idea still belongs more to the experimental realm than to real-world implementation, exploring it helps us better understand where the advertising infrastructure of streaming could evolve.

The Current Model: Inserting Segments into a Segmented Stream

To understand why these ideas are emerging, we first need to understand how modern streaming works.

Dominant protocols such as HTTP Live Streaming and MPEG-DASH are based on three fundamental principles:

1. Video segmentation into small chunks lasting a few seconds.
2. Manifests that indicate which segments the player should request.
3. Distribution through CDNs to scale global traffic.

When an ad break occurs, what actually happens is that the playback stream switches to another set of segments containing the advertisement. This process is known as stitching.

Depending on where the ad decision is made, we refer to:

• CSAI (Client-Side Ad Insertion): The player requests the ad and manages playback.
• SSAI (Server-Side Ad Insertion): A server generates a personalized manifest in which the ad segments are already integrated.

This model has enabled advertising to scale within streaming environments, but it also introduces certain technical limitations.

The Limitations of the Current Model

People often talk about an alleged “glass ceiling” for CSAI and SSAI. In reality, the problem lies less in the model itself and more in the architecture of modern streaming.

In the case of CSAI, the challenges are well known:

• player dependency
• device fragmentation
• issues with ad blockers
• measurement difficulties

For this reason, many OTT services have migrated toward SSAI, which offers several advantages:

• greater control over the video stream
• compatibility with CTV devices
• better integration with streaming infrastructure
• more reliable server-side measurement

However, even within the SSAI model some technical challenges remain:

• small quality variations between content and ads
• latency during transitions between segments
• complexity in managing personalized manifests

The industry has addressed part of these issues through techniques such as:

• encoding alignment between content and ads
• frame-accurate ad markers based on signaling
• prefetching of ad segments

In practice, these advances have already made the advertising experience on many OTT services relatively seamless.

But there is a deeper structural limitation.

The Real Limit: The Base Content Does Not Change

The current model allows us to personalize which ad is shown, but it does not allow us to modify the video content the user is watching.

For example, today we can decide:

• which ad is shown
• to which user
• at which moment within the content

But we cannot modify elements inside the video itself, such as:

• changing a brand that appears in a scene
• altering a billboard inside a stadium
• replacing products within a shot

This is where new experimental approaches such as in-scene advertising or virtual product placement begin to appear.

Some companies are already developing technologies capable of inserting brands directly within audiovisual content.

This approach does not replace traditional ad insertion, but it opens the door to a new form of advertising that is more integrated into the content itself.

The Growing Role of Edge Computing

Another relevant evolution within the ecosystem is the growing role of edge computing in distribution infrastructures.

Major CDNs such as Fastly, Cloudflare, and Akamai are incorporating code execution capabilities closer to the end user.

This allows part of the processing logic to move away from centralized servers, including:

• content personalization
• manifest manipulation
• advertising decision logic
• delivery path optimization

In this context, some have proposed a conceptual evolution: if we are already personalizing streaming at the edge, why not generate or modify ads directly there?

From “Stitching” to “Composition”

A more radical idea is to treat video as a composition of layers: the content becomes the base layer, and ads are dynamically generated as additional layers.

Technologies such as WebAssembly allow high-performance code execution at the edge. However, rendering video in real time for millions of users remains costly and complex, particularly due to the requirements of decoding, frame modification, and re-encoding. It would also break the efficiency of CDN caching.

Looking Ahead: The Advances Required

For these types of models to become viable, several technological advances would be necessary:

• Richer signaling based on standards such as SCTE-35, capable of carrying not only ad break markers but also rendering parameters.
• Video processing capabilities at the edge within CDNs that allow efficient modification of the stream.
• Object-based composition techniques within video, separating foreground and background in real time.
• A significant reduction in the cost of cloud-based graphics processing through GPU-as-a-Service models.

These elements could make advertising generated “at the source” viable, rather than relying on segmented and pre-encoded ads.

More Evolution Than Revolution

It is unlikely that, in the short term, advertising will stop being video inserted into a video stream.

However, we are already seeing a gradual evolution in three directions:

1. Increasingly sophisticated SSAI, with better signaling, lower latency, and deeper personalization.
2. New advertising formats embedded within content, such as virtual product placement or scene-level contextual advertising.
3. More logic executed within the distribution network, enabled by edge computing and increasingly programmable infrastructures.

Rather than replacing ad insertion, these innovations will likely lead to hybrid models where multiple forms of advertising coexist within the same streaming ecosystem.