“A few years ago, as repurposed short form content became more and more widespread, increasingly powerful large language models and advanced transcription tools started to emerge. We realized we could use them to automate the most tedious aspects of the editing process,” explains Champion, President and Co-founder of Zigg. Users upload their content to Klap’s web interface, where AI models use the audio to detect engaging “short-worthy” moments and automatically reframe footage from landscape to portrait. Within 10 minutes, they receive a gallery of short, vertical videos ready to download, publish, or fine-tune in the platform’s built-in editor.

Each new spike in users required manual adjustments and added infrastructure complexity.

The company decided to migrate video processing, storage, and task management into one unified environment. Now, Compute Engine handles the heavy lifting of video transcoding and AI processing, while Cloud Storage stores the platform’s growing library of user videos. Pub/Sub manages the flow of distributed processing tasks, ensuring each job is queued and executed efficiently. “This setup has enabled us to lower costs significantly,” says Champion. “Since consolidating on Google Cloud, our storage and processing costs are a third of what they would have been elsewhere.”

Faster processing, fewer errors, and greater uptime have opened the door to new enterprise contracts, where trust in infrastructure is a key part of the sales process. Zigg is preparing to scale its B2B offering and serve more enterprise clients through its API model. “When clients see we’re built entirely on Google Cloud, it gives them confidence that we’re serious and scalable,” says Champion. “Reliability is key for enterprise users — they require 99.9% uptime. With Google Cloud, we’ve seen huge improvements in that area.”

Looking ahead, Klap is expanding its AI capabilities beyond speech-based editing, exploring multimodal models that understand both audio and visual context and recognize not just what’s said, but what happens on screen. This will demand significant GPU power, large-scale data management, and efficient model training pipelines.“Right now, we’re focused on compute and storage,” says Champion. “But as we train larger models, we’ll explore Vertex AI to handle future workloads.”

“For me, it all comes down to control, cost efficiency, and the kind of reliability that helps me sleep at night,” he concludes. “Google Cloud gives me all three.”