Universal 3 mejora la transcripción en ruido real (Full Transcript)
[00:00:00] Speaker 1: Ryan, we just launched Universal 3 with our CallRail customers for our voice assist product. Like, why is it so good?
[00:00:06] Speaker 2: Yeah, Universal 3 is really robust to real-world scenarios that actually break in voice. It's great at- You gotta speak up, I can't hear you. Oh yeah, it's great at names, emails, phone numbers, alphanumerics. So if you say something like, Lewis Hamilton at Ferrari, he's number 44, right? He's from Great Britain, he's one of the best drivers in the world. You'll get great transcription on that. But it's also amazing if you say, you know, my email is rseams, r-s-e-a-m-s, at assembly.ai.com, you're gonna get those email addresses as well. And so in these real-world conditions, like we are right now with the cars driving behind us here in Miami, you're gonna get awesome transcription.
[00:00:43] Speaker 1: Yeah, what happens when, like, Kenny Antonelli passes, like, Max Verstappen when they come by? Will it pick it up? It should, but let's test it and see. Yeah.
- Universal 3 está diseñado para resistir condiciones reales que suelen romper el reconocimiento de voz.
- Maneja bien entidades difíciles: nombres, correos electrónicos, teléfonos y alfanuméricos.
- La precisión se mantiene incluso con ruido de fondo (p. ej., coches pasando).
- Ejemplos prácticos muestran buen desempeño con deletreo y datos estructurados.
- Planean validar el rendimiento con una prueba en vivo cuando pasen coches/pilotos.