How do you fill the gaps in programming left by shooting schedules wrecked by Covid-19? Arrow International Media turned to artificial intelligence (AI) to mine its archive – and create new shows.
In late August, RTS London invited a panel of Arrow representatives, chaired by Muki Kulhan, to explain how the factual indie did it.
Production executive Carrie Pennifer explained that lockdown had meant no shooting or access to the edit suite, and everyone working remotely. Post-production manager Kyran Speirs had more than 20 unfinished programmes to deliver.
A cross-company team brainstormed solutions and turned to the company’s cloud-based digital archive for suitable content, but it was not fully logged. To do this manually would take months, realised post-production consultant Dan Carew-Jones – at which point, the company picked up on previous conversations with GrayMeta about AI.
GrayMeta’s Curio system could tap into Arrow’s digitised content immediately, providing visual tagging, optical character and speech recognition, and camera metadata extraction. Carew-Jones gave an example of a search for a policeman, where the system was able to recognise his police badge.
Runner-turned-logger India Goss found that, whereas previously she was laboriously logging shots manually, she was now able to find relevant footage via Curio to offer to the editor. The AI wasn’t perfect, but it helped to harvest 2,000 hours of usable content in a few days.
The Arrow team has learned a lot, not least the value of digitising rushes and capturing camera metadata.