United KingdomMedia and entertainment2024 – ongoingAudiomob
AI review pipeline for audio and image ad creatives
Audiomob cut manual ad reviews from 20,000 to about 70 a day with an AI review pipeline
Problem
Audiomob's ad platform handled more than 20,000 audio and image creatives a day. The company had to keep inappropriate content away from players, and reviewing every creative by hand was not possible.
Approach
Audiomob built the pipeline in two weeks from Google Cloud's Speech-to-Text, Natural Language and Vision AI services, so it did not train its own models. Audio ads are transcribed and image ads are read with optical character recognition, and the text is then categorized and checked for offensive material. Google SafeSearch checks the images that come with audio ads. The tools give confidence levels per category, and the creatives most likely to be prohibited go to manual review.
Results
Takeaway
Ready-made speech, text and image services let a 33-person company screen ad content at a volume that manual review could not cover. Staff only look at the creatives the pipeline flags.
Categories
- Size
- Small business
- Industry
- Media and entertainment
- Business function
- Production and quality
- Innovation type
- Process innovation
- AI technology
- Computer visionPredictive analytics and ML
- Adoption
- Off-the-shelf tool or platform
- Evidence
- Vendor case study
Sources
Source summaries
Google Cloud's customer story describes how Audiomob, a UK mobile-game audio advertising company, built an AI pipeline to review the more than 20,000 ad creatives it receives each day. The story gives manual reviews falling from 20,000 to about 70 a day and a pipeline 100 times cheaper than manual review. The page carries no date; the Internet Archive first saved it on June 13, 2024.
Full details are in the original sources.
