Otter branding

Lost in transcription: the battle for AI accuracy in converting speech

Of the many transcription platforms out there, Otter sits head and shoulders above the rest for a decent voice to text transcript. But despite its valid functionality claims, those needing good tools in this space are often still frustrated

If you’ve ever tried to use any form of AI-powered speech to text transcription, you’ll likely be aware that for the most part, it can be shonky at best. The peculiar turns of phrase they generate can often be hilarious (see also: YouTube auto generated captions on regional accents); occasionally they can even seem poetic, but most of the time, the swathes of inaccuracies and nonsense are just plain irritating. The closest thing I’ve found to something vaguely useful when it comes to transcription is a site called Otter, which also wins points for its name’s cute factor. 

However, Otter is keen to stake its claim as a lot more than just a transcription service, as a recent overhaul of the platform which positions it firmly as a ‘productivity tool’ drives home.

The redesign was helmed by Otter chief design officer Richard Ward, who joined the company in its San Francisco office around a year ago. Ward moved to the US from the UK 15 years ago, taking in stints at Frog Design, Ideo and West Coast agency MetaLab, which created the original Otter branding when the platform launched in 2017.

Over the past five years, Ward says that the “core DNA” of Otter has shifted, and this seems to be largely related to the explosion in home and hybrid working catalysed by the Covid-19 lockdowns. He acknowledges that the “original kernel of the idea” around creating an AI-powered transcription service remains, but Otter is now far more about how its tools can “drive collaboration” through the content generated in live audio transcription. “We’re able to do a lot with it: tag it, highlight it, comment on it, create action items — we’re able to basically drive collaboration in that live meeting with multiple participants,” Ward explains.