61–70 of 73 posts
[flagged]
This looks really interesting. I feel like there is a lot of room to build great voice-based agents that don't exist right now. I have found that ChatGPT voice mode is unusable (e.g. hallucinates me saying things); Claude voice mode is usable, but very buggy around tool calling, and it often mishears things. And it only supports Opus, not Fable (though it looks like you don't support either of those). But I use it an…
I agree. Even Codex’s new voice mode isn’t just “It does things I tell it”, all commercially available voice modes have to be dumbed down in the name of latency for some reason. I have even experienced the latency being uncomforably low before.
This is what I enjoy the most about my in-development personal OpenCode client: It’s just STT-LLM-TTS with the original model and all of its capabilities. Latency is obviously higher, but I don’t mind that much given the upside (would be interesting to optimize for completely blind usage, my current MO is mostly hybrid).
[flagged]
i don't have a thesis on voice as a form factor for search. the demand we serve already exists: businesses answer phones. receptionists, outbound campaigns, clinic front desks and these calls happen at scale today, and the teams running them are the ones picking speech models. whether voice wins new interfaces is a separate bet
Does this include a turn taking API? It'd be great to have one API that could do "Conversation in a box". One of the biggest annoyances is daisy chaining many models together for turn taking, dumb models for immediate responses, with smarter models returning and taking over after.
Yes, on the hosted side (agents platform): full sessions come with VAD and turn-taking handled - we set them up and tune them for your use case, so that is the closest thing to conversation in a box. If you run your own orchestration, the gateway is just the routing layer and turn-taking stays in your framework - in our own stack we run Pipecat's Smart Turn in-process and tune the commit threshold on real calls. We a…
Earlier quoted context omitted.
i don't have a thesis on voice as a form factor for search. the demand we serve already exists: businesses answer phones. receptionists, outbound campaigns, clinic front desks and these calls happen at scale today, and the teams running them are the ones picking speech models. whether voice wins new interfaces is a separate bet
So how exactly are you counter-positioned to Eleven Labs? They are serving a similar market correct?
Earlier quoted context omitted.
So how exactly are you counter-positioned to Eleven Labs? They are serving a similar market correct?
not really counter-positioned, they are one of the providers we route to. elevenlabs sells models - voices, stt, and their agent platform on top of them. we don't sell any models: we measure all of them (elevenlabs included, their scribe is near the top of our stt board) and route each request to whatever wins for your language and constraints. so for them the best outcome is that you use their models; for us the bes…
Does the optimizer model cache-hit rate and context distribution, or does it score on list rates?