Kugel Ear is a speech-to-text model from KugelAudio, a voice-AI company whose models are developed and hosted in Europe. Its distinguishing behavior: when it is unsure of a word - a caller's name, say - it tells the bot and supplies the spellings it heard, so the bot asks once instead of guessing.
The vendor publishes a benchmark table comparing Kugel Ear 1 against ElevenLabs Scribe v2, OpenAI gpt-transcribe, Speechmatics Enhanced and Deepgram Nova-3 on four tests; those numbers are the vendor's own.
Broader platform claims, attributed to the vendor: natural voices in 39 languages, GDPR compliance, EU sovereign cloud included on every plan for EU/EEA customers, and first audio in 91 ms median on its Kugel 3 model, shown with a sample call transcript.
