Viewing profile — abdik

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: …

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 …

thanks! yes, some of them do, cuz speed is a per-provider capability, not universal. And, you can see it in the gateway code (minimax, hume, xai tts adapters all handle a speed par…

People who built or building voice agents immediately get this. openrouter is the openrouter for audio models. the conflation is "audio models" vs "voice ai", and the mental model …

yes, it does support, when you are creating an api key, you can point out narration or transcription use case, then you will be able to see. let me know how it goes or what use cas…

looks cool, checking it out!

exactly, we see the same thing, around 95% cases are still cascaded, even tho STS has been improving a lot

Synthetic-data fine-tuning is the other credible answer to domain vocabulary. Curious whether you re-benchmark the fine-tune when new base models ship?

What purrcat259 said, and I think that the page should explain it, we will add a tooltip. for some languages CER is more relevant than WER. Thai and Mandarin have no word boundarie…

agree with omneity here. Whisper's initial-prompt trick is exactly that, and several hosted vendors have equivalents (custom vocabulary / keyword prompting). Domain vocabulary is w…

Actually, we measured exactly this recently. The strongest open-weights speech-to-speech model we have run is NVIDIA's NemotronLabs VoiceChat 11B - no provider serves it, so we hos…

cool app, and agreed that on-device keeps eating the single-user cases, we are seeing dictation and translation are exactly where local models shine. We benchmark the open models o…

Turn-taking specifically does not need a listening panel, but it is measured mechanically. 200+ real human clips, and we score end-vs-wait decisions: did the model decide the calle…

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 con…

there is a good progress on on-devise models, but not ready for production yet to fit in devices. But as soon as there is are some good results, we are going to benchmark them and …

thanks! actually, we have the filipino already, can you check out and share your feedback?

Fair pushback. On end to end: we measure those too, same methodology: https://benchmarks.speko.ai/s2s . If the single models win, we route to them the same way, so we do not care w…

The main difference from gateway is we help with picking the right voice stack, which seems to be a big problem for users: we benchmark the models continuously and route based on t…

Thank you!

Yes, i added it. that's the right link.

Hi HN! I'm Bek, founder of Speko, a platform that finds an optimal combination of speech-to-text, LLM, and text-to-speech models, given your constraints, among all our public bench…

Got into YC interview recently and built this to practice, it's been helpful. Sharing, so it can be used by others well. Good luck everyone!

The logo is similar to ours https://www.lovo.ai/

Check out https://nomadlist.com/