Otter adds 13 transcription languages, with automatic language switching still to come

Otter users can now record meeting notes in 13 more languages, but letting speakers switch languages without changing a setting will have to wait. In its November 10 announcement at Web Summit Lisbon, the company outlined a staged expansion to 31 languages, pairing immediate access with a forthcoming option for following mixed-language conversations.

More languages today, fewer setting changes later

Before this release, Otter supported English, French, Spanish and Japanese. The 13 additions take its current coverage to 17 languages. The company describes this as its largest language expansion.

• Available now: Mandarin Chinese (Simplified), German, Italian, Cantonese, Korean, Brazilian Portuguese, Russian, Indonesian, Malay, Dutch, Turkish, Vietnamese and Polish.
• Coming soon: Swedish, Danish, Finnish, Arabic, Hindi, Thai, Hebrew, Norwegian, Greek, Macedonian, Persian, Czech, Hungarian and Romanian.

For now, users choose a supported language in Otter’s settings. The forthcoming multilingual option is intended to remove that manual step when participants move between supported languages during the same meeting. Otter gives the example of a sales call moving from English to Portuguese to Mandarin and back.

That distinction matters for anyone choosing the tool for a multilingual team: support for a language and automatic handling of language changes are separate capabilities. Otter labels the latter “coming soon,” rather than available with today’s additions. Its announcement gives no release date for that option or the remaining languages.

The expansion reaches beyond the transcript

The added languages come at no additional cost across Otter plans and are available on desktop, mobile and browser.

The announcement also says Otter’s existing Work Mode can turn conversations into follow-up emails and customer relationship management updates in the meeting’s language. Those are language extensions to the surrounding workflow, not a separate product launch: the change announced here is the set of languages that workflow can serve.

In-house models, an English accuracy figure

Otter emphasizes that it trains, tunes and runs its own speech models on its own infrastructure, rather than wrapping a third-party speech service. It also says it does not send audio to third parties for processing, and that this commitment extends to each new language as it becomes available.

The company told TechCrunch that the newly added languages use models trained on publicly available datasets. Native-speaking volunteers reviewed transcripts and checked accuracy. That describes a development and review process, while the numerical benchmark highlighted in Otter’s own announcement concerns English: it reports average accuracy of 96% on the Open ASR Leaderboard.

That English result is not a measure of accuracy in the newly added languages. Otter’s release supplies no equivalent language-by-language results for those additions. The launch establishes broader access; it leaves users without the same numerical evidence for judging transcription quality across the expanded language list.