Kugel Ear asks back when it's unsure, so your callers never have to spell out their name again.
 | Introducing Kugel Ear | Another week, another release, today we’re excited to announce Kugel Ear! Our speech-to-text model built for phone calls, specifically trained for customer-support use cases, ahead of ElevenLabs, Speechmatics, OpenAI and Deepgram. When Kugel Ear isn't sure about a word, it asks, the way a person would: “Elizabeth with a Z or S?” Just that one word, never the whole name letter by letter. Your callers don't have to spell out their name anymore. | | Kugel Ear at a glance |  | Why build on Kugel Ear | | • | No more spelling loops. Unsure words reach your LLM flagged, with the other spellings Kugel Ear heard, so your bot asks “Philip with one L or two?” instead of saving the wrong name. It flagged 86% of the words it got wrong. | | • | Built for phone calls. 93.4% of words right on customer-support calls across 14 accents, 88.5% even through a phone line, and the final transcript is on its way 72 ms after the caller stops. |
| One more thing | Like every other release, we're live on Linkedin under this Link. Product Hunt is coming tomorrow (leave a like and a comment, possibly even a repost 👀). | | | You receive this email because you subscribed to “Product updates” at KugelAudio. Unsubscribe · Account settings KugelAudio GmbH · Rosenthaler Str. 36 · 10178 Berlin · Germany
Register: Amtsgericht Charlottenburg, HRB 277989 B · VAT ID: DE458999009
Managing directors: Kajo Kratzenstein, Viktor Presber |
|
|