Generative AI: Sportsnet Drop-in Ad for Raptors
A broadcast-ready TV drop-in ad — built with AI in days, not weeks.
Context
Live sports broadcasts create moments of high viewer attention during pauses in play. This concept explored how generative AI could enable brands to rapidly produce animated mascot content designed specifically for pause-ad placements during NHL broadcasts.
Developed as exploratory work for belairdirect, the idea focused on transforming idle broadcast moments into playful brand interactions using AI-assisted animation.
The Idea
Pause ads offer a unique opportunity for brands to appear directly within the broadcast experience. Instead of static placements, the concept imagined a 3D brand mascot reacting dynamically to game moments — celebrating goals, reacting to penalties, or engaging viewers with short branded animations.
Using generative tools, the goal was to prototype a workflow that could quickly generate these animations without requiring a full traditional 3D production pipeline.
AI Workflow
AI Tools Used / Tested
- ◆Claude
- ◆ChatGPT
- ◆ElevenLabs
- ◆Veo (Google Gemini)
- ◆Seed Dance
- ◆Adobe Firefly
Adobe Suite
- ◆After Effects
- ◆Photoshop
- ◆Illustrator
Process
- ◆Concept exploration and scripting using AI assistants
- ◆Voice generation for mascot dialogue using AI voice synthesis
- ◆Generative video exploration to animate the mascot and environments
- ◆Motion refinement and compositing
- ◆Final prototype demonstrating pause-ad behaviour and animation style
The Static Hero Pose
The belairdirect mascot as a transparent PNG — the source asset fed directly into Veo as the subject of the animation.
The Prompt Given to AI
In order to get a detailed prompt for Veo, I used ChatGPT & Claude to design and develop the prompt such that it could be directly injected into Veo:
“Act like a Creative Director and Motion Designer. I need a prompt to give to Gemini Veo to animate the transparent PNG I just shared. The motion sequence is the mascot is standing on the benches, watching a basketball match. He is jumping up and down in excitement as he sees a basket being scored. I need the video to be 10s long and in 16:9 format. Make sure none of the features of the mascot are disrupted or it doesn't look like he's bending out of shape. Make sure to keep its original colours.”
The Result
The following prompt was generated by Claude and ChatGPT, refined for direct injection into Veo:
“3D animated cartoon mascot standing courtside at a professional basketball arena during a live game, Toronto Raptors-style arena environment, crowd visible in the background and bright arena lighting illuminating the court, a basketball shot is made and the ball goes through the hoop, the mascot watches the basket being scored and immediately reacts with excitement, the mascot jumps up and down repeatedly in celebration, energetic bouncing movement, pumping its arms in excitement, big expressive facial reaction, joyful and playful cartoon energy, the basketball hoop and backboard visible in the background, crowd cheering atmosphere, camera framed like a live sports broadcast cutaway, medium shot focused on the mascot reacting to the basket, slight handheld broadcast camera feel, high-quality 3D animated commercial style, family-friendly sports entertainment tone, short 5-second animation designed for Sportsnet broadcast drop-in pause ads.”
Veo 3 Test Renders
The Issues
Working within the current limitations of generative video tools, three key issues emerged and had to be resolved:
- ◆The output of the video was unable to be produced in 16:9 aspect ratio
- ◆The output video was unable to exceed 7 seconds — since this was supposed to be a one-shot animation, the entire 10-second render was important
- ◆The output was unable to be a transparent background video. In order to fix that, we added a green layer behind the original transparent PNG so that we could key out the background on the final render
Final Production
Building the Scene
In order to stitch the whole video together, we used Adobe Firefly and Adobe Stock to create the scene. We downloaded an image from Adobe Stock that had VIP seats that could be used for a basketball stadium — these seats were where the mascot would be jumping up and down.
Extending the Environment
We then used Adobe Firefly and Adobe Photoshop to complete the scene by prompting it to fill the left and right of the seats with seats from the stadium, so that we could have a complete environment.
Compositing in After Effects
Now that we had a scene ready, we used Curves and Levels in Photoshop to adjust the background so that the focus remained on the VIP seats and not on the background seats. We then imported the whole composition into After Effects — placing the final AI-rendered video from Seed Dance, adding shadows to make the assets belong to the scenario, adding the other animations, and colour correcting and grading the entire sequence for final production.
Basketball Players & Ball
To create the basketball players, we downloaded a stock video of two players on a court. We keyed out their background and using a displacement map stretched them out to make them look like shadows playing on the court. We used simple keyframes to animate the basketball and matched the timing with the end of the basketball clip.
Voice Over — ElevenLabs
Finally, we needed the VO to sync perfectly with the animation of the end sequence: "Belairdirect. Insurance. Simplified." In order to facilitate that, we leveraged ElevenLabs — generating a VO with a script and sufficient pauses as a guide for the animation. This same VO and script guide was delivered to Sportsnet for their VO artist to match.
Final Outcome
Broadcast-Ready. Built with AI.
In the end, we generated a high-polished, 3D Animated TV Drop-In ad for the client using Generative AI tools, which would have cost the client weeks to produce at around $10,000–$20,000. This ad — along with others — aired on Sportsnet during the Raptors game interval and received much praise from the internal and external team from belairdirect.
