How SDVI, Vionlabs, and Codemill turned a 20 to 60 minute manual process into a minutes-long, AI-powered workflow, with a human still in the loop where it matters.
Ad break detection is one of the most labor-intensive steps in preparing content for monetization: slow, manual, and spread across disconnected tools. Meanwhile, ad-supported tiers are the fastest-growing segment of streaming, and viewer tolerance for bad ad breaks is zero. In this webinar, SDVI, Vionlabs, and Codemill show how their integrated solution combines story-aware AI detection with frame-accurate human validation, all orchestrated inside a single media supply chain, to cut ad break workflows from hours to minutes without sacrificing quality.
46% lower viewer churn when ad breaks are placed at story-aware, natural pauses.
What you'll learn
- Why ad break detection is a scaling problem, not just a workflow inconvenience
- How Vionlabs' story-aware AI identifies natural pause points, not just scene cuts
- How Codemill's Validate tool puts a human in the loop for frame-accurate review, with keyboard-driven efficiency built for high-volume teams
- How SDVI's Rally platform orchestrates the entire process end to end, from ingest to SCTE markers, scaling resources on demand
- A live look at how AI-generated markers compare to human judgment, including where AI can get it wrong
Speakers
- Geoff Stedman, Chief Marketing Officer, SDVI
- Arash Pendari, Founder & CEO, Vionlabs
- Neil Anderson, Chief Revenue Officer, Codemill
- Chris Brähler, Chief Product Officer, SDVI
Watch now to see the full solution in action, and learn how it can map to your own ad-supported content pipeline.
Webinar Summary
The problem
Ad break detection sounds simple, but it's one of the most labor-intensive steps in preparing content for ad-supported monetization. A single 40-minute episode can take 20 to 60 minutes to segment manually, and most teams rely on a fragmented mix of spreadsheets, players, and MAM systems with no single source of truth. That doesn't scale: content libraries are growing several times faster than the teams supporting them, and ad-supported tiers (FAST, AVOD, etc.) are the fastest-growing segment of streaming.
The stakes are real money, not just workflow annoyance. Poorly placed ad breaks, especially ones that land mid-scene or mid-dialogue, measurably increase viewer churn and reduce ad value. Placing breaks at story-aware, natural pauses instead cuts viewer churn by 46%.
The solution: three companies, one workflow
SDVI, Vionlabs, and Codemill have built an integrated, cloud-native solution that automates the repetitive parts of ad break detection while keeping a human in the loop for the judgment calls. It works in four steps:
- Orchestration (SDVI Rally): Content lands in Rally, which kicks off the workflow, provisions resources, and routes the asset to Vionlabs automatically, with no manual handoffs.
- AI detection (Vionlabs): Vionlabs' story-aware AI analyzes the content and returns candidate ad break markers in minutes. Rather than just detecting scene cuts, it looks at narrative structure: scene changes, black frames, audio silence, and shifts in sentiment between scenes, to rank the best possible ad break locations.
- Human validation (Codemill Validate): Rally sends the proxy and candidate markers to Codemill's Validate tool, where an operator reviews, adjusts, and approves the final markers on a frame-accurate timeline. Keyboard shortcuts and a single unified interface (instead of jumping between three separate tools) are built specifically to speed this step up for teams doing this at high volume.
- Delivery: Once approved, the finalized markers flow back into Rally, which drives the asset through packaging and ad stitching, delivering markers as SCTE markers or sidecar files to downstream ad insertion systems.
Rally acts as the connective tissue across the whole process. The operator only touches the asset for the high-judgment work; everything else runs automatically.
Why this approach works
The solution combines best-of-breed applications, drawing on Vionlabs and Codemill's specific expertise, rather than trying to do everything in one platform. It's cloud-native and consumption-based, so teams can scale resources up for a large library backfill over a weekend and scale back down once it's done, paying only for what they use.
On AI accuracy (Q&A highlight)
Asked how the AI-generated markers compare to what an experienced human editor would choose, Vionlabs' Arash Pendari was candid: a rested, focused human can do a very good job, but AI wins on speed and consistency, and it isn't infallible. He gave a concrete example: a camera panning across a wall into darkness can be misread by the model as a black frame and mistaken for an ad break point. This is exactly why human validation remains a core part of the workflow, not an optional add-on, and why his advice for anyone using AI models (regardless of vendor) is to always double-check outputs and keep a human in the loop.
On operational efficiency (Q&A highlight)
Codemill's Chris Brähler illustrated why small efficiency gains compound at scale: if operators need to hunt across three different tools to find the right content and get set up, that alone can cost 5 to 6 minutes per task. Multiply that across 10 tasks a day and 10 people, and you lose 10 hours a day just to navigation, time that could otherwise go into actually processing ad breaks. Having everything packaged into a single "work order" that launches directly into Validate removes that friction.
Bottom line
The solution is live in production today. It's built for teams handling library backfills, new FAST channel launches, or anyone trying to get more value out of an ad-supported tier, and the three companies are offering to map it to a viewer's specific pipeline and workflow.
.png)
