Kugel Ear

Speech-to-text for voice agents that flags words it is unsure of and sends the alternative spellings heard, so the bot asks once instead of making callers spell everything out

Paid Video & Audio
Visit Website ↗ by KugelAudio

What is Kugel Ear?

Kugel Ear 1 is speech-to-text built for voice agents: unsure words reach your LLM flagged, with the other spellings it heard and their probabilities, so your bot asks once instead of making callers spell. A pause inside a sentence is not treated as the end of a turn, so the bot does not talk over a caller reading out a number.

On AppTek's public call-center dataset (280 English customer-support turns across 14 accents), Kugel Ear 1 got 93.4% of words right on clean audio and 88.5% through a simulated phone line, with the final transcript delivered in a 72 ms median. One WebSocket carries audio in and the transcript plus alternatives out, across nine European languages, at €0.48 per hour excluding VAT.

Key features

  • Uncertainty flagging with alternative spellings and probabilities
  • Smart end-of-turn detection that waits out mid-sentence pauses
  • 72 ms median time from end of speech to final transcript
  • One WebSocket API: audio in, transcript and alternatives out
  • Nine European languages with per-turn language detection
  • Top of its maker's 16 October 2026 benchmark against ElevenLabs, OpenAI, Speechmatics and Deepgram

Pros

  • Kills spelling loops on names, addresses and customer numbers
  • Fastest final transcript in its published benchmark

Cons

  • No free tier is mentioned on the product page

Kugel Ear at a glance

PricingPaid
CategoryVideo & Audio
Platformsweb
CompanyKugelAudio
Best forVoice-agent builders whose bots take names, addresses and customer numbers over the phone
Official websitewww.kugelaudio.com
speech-to-textvoice agentstranscriptioncall centerllm

How to choose the right Video & Audio tool

Video and audio tools split into generators (which create footage, voices or music from prompts) and editors (which speed up cutting, captions and cleanup). Start from your end product: short social clips, long videos, podcasts or voiceovers each reward different strengths. For generation, check output resolution, clip length, and how much control you get over style and consistency. For voices and music, listen for natural pacing and check the languages you need. Licensing is critical — confirm you can use the output commercially, especially for client work and monetised channels. Finally, weigh render times and credit costs: a tool that takes an hour per clip changes your whole workflow.

Compare all Video & Audio tools →

Kugel Ear FAQs

What is Kugel Ear?

Kugel Ear 1 is speech-to-text built for voice agents: unsure words reach your LLM flagged, with the other spellings it heard and their probabilities, so your bot asks once instead of making callers spell. A pause inside a sentence is not treated as the end of a turn, so the bot does not talk over a caller reading out a number.

What are the main features of Kugel Ear?

Kugel Ear offers Uncertainty flagging with alternative spellings and probabilities, Smart end-of-turn detection that waits out mid-sentence pauses, 72 ms median time from end of speech to final transcript, One WebSocket API: audio in, transcript and alternatives out, Nine European languages with per-turn language detection and Top of its maker's 16 October 2026 benchmark against ElevenLabs, OpenAI, Speechmatics and Deepgram.

How much does Kugel Ear cost?

Kugel Ear is a paid product. Prices and plans change over time, so check its official website for the current pricing before you buy.

What are the pros and cons of Kugel Ear?

Strengths of Kugel Ear include Kills spelling loops on names, addresses and customer numbers; Fastest final transcript in its published benchmark. On the downside, No free tier is mentioned on the product page.