Loader ads.
Ads during the wait, separate from the result. Loader ads appear while an AI task generates its output. They use an existing wait to help fund compute without changing the model’s answer.
Especially suited to video, image and music generation, where processing time creates a natural window for a separate ad.
wavebird’s mechanism for placing ads during AI generation is patent pending.
The user asks for a result.
A separate ad can appear while the model works.
The AI task finishes independently.
No ad ready? Continue the AI task.
Fund the compute. Keep the answer independent.
A distinct visual placement
Brand, imagery and a clear sponsored label distinguish the ad from an AI recommendation or factual answer.
Your product controls the moment
Keep model output and ad delivery on separate paths. Choose timing and rendering, without delaying generation.
One format. A wider set of possibilities.
Loader ads
A labeled visual ad during existing AI generation time. The result stays independent.
Answer-adjacent ads
A labeled visual ad beside or below the completed answer, outside the model’s words.
Rewarded compute
Proposed formatThe user chooses an ad in exchange for additional AI usage.
What could the reward be?
Another generation, additional tokens, more images, temporary premium-model access, longer context or higher-resolution output.
Sponsored compute
Proposed formatA sponsor funds a defined generation, job, token allocation or product moment. The output remains independent.
Link the ad to the generation.
Associate a valid ad exposure with the AI task it accompanied. The generation alone does not prove an ad was seen, and a longer wait alone does not establish additional billable impressions.
This follows the Compute Sponsoring proposal for linking exposure to identifiable compute activity.