OpenAI GPT-Realtime-Translate debuts live translation

This is the useful version of voice AI: not just cleaner post-production dubbing, but low-latency translation that can keep up with live speech. If the fluency holds up outside demos, OpenAI just turned multilingual video and events into a product feature instead of a separate workflow. Vimeo’s demo is a practical proof point, not just a lab demo. The 70+ input / 13 output split is strong on ingestion, but output coverage still limits full localization. At $0.034 per minute, OpenAI is pricing for scale testing and deployment, not artisanal dubbing. The real competition is specialist video-translation vendors and event platforms, not just other model APIs.