Show HN: Open source framework OpenAI uses for Advanced Voice

Imagine being able to tell an app to call the IRS during the day, endure the on-hold wait times, then ask the question to the IRS rep and log the answer. Then deliver the answer when you get home.

Or, have the app call a pharmacy every month to refill prescriptions. For some drugs, the pharmacy requires a manual phone call to refill which gets very annoying.

So many use cases for this.

As costs of humanlike communications decrease, so will Sybil attacks and spam.

The IRS is notorious for resistance to tech change, don't be surprised if they unplug the phones and force you to walk in to ask your question.

What is the value add here? Save sometime for technocrats and technoadjacents for a whole of 3 years before victims of spam adapt?

Also this has been solved already just mail your question like the rest of mortals.

It would be really nice if the IRS would ALLOW you to walk in and ask a question!

That is very expensive. Offices all around the country with personnel. We are going to have to fund them instead of gripe about them to get that to happen.

Yeah, that is why I doubt it will happen. Maybe a website where you can sumbit an issue and have it resolved in a reasonable number of days would be fine.

Years ago my tax return was flagged as a possible fraud case -- I believe a direct consequence of a big data breach. I had to go into my "local" IRS office and present my passport to prove indeed it was me. Decidedly not nice.

True to form, with an appointment I waited 3 hours at the office and watched the guard staff turn away countless people. Finally saw a person, gave then my passport, and finished in a minute.

I am going through that right now. IRS owes me 3 years of refunds, but I can't even get an appointment to see them. They hang up one when I call (after hours on hold), and won't let me just visit the local office. My current attempt is to work with my US Senators office.

they would make you wait for 3 hours queueing

They'll just put up "captchas" or whatever.

The point if phone lines is to waste the client's time. Not to have the client waste their time.

We do this exact thing at getvibrato.com. You can schedule calls like these, or even do more advanced automation with Zapier.

This is really helpful, thanks!

OpenAI hired the ex fractional CTO of LiveKit, who created Pion, a popular WebRTC library/tool.

I'd expect OpenAI to migrate off of LiveKit within 6 months. LiveKit is too expensive. Also, WebRTC is hard, and OpenAI now being a less open company will want to keep improvements to itself.

Not affiliated with any competitors, but I did work at a PaaS company similar to LiveKit but used Websockets instead.

> LiveKit is too expensive

Most of it is open source, especially the clients, although they do feel quite ad hoc hacked together (a possible side effect of WebRTC evolution).

Would totally agree on OpenAI moving away. The description of the agent here sounds like a big hack just to get around the fact temporarily the model server expects audio over sockets instead.

Which components feel ad hoc?

In most real applications, the agent has additional logic (function calling, RAG, etc) than simply relaying a stream to the model server. In those cases, you want it to be a separate service/component that can be independently scaled.

Essentially I think the Livekit value is a SFU that works, with signalling, and the SDKs exist. My experience is people radically overstate how hard signalling is, and underestimate SFU complexity, especially with fast failover.

In terms of being a higher level API arguably it is doomed to failure, thanks to the madness of the domain. (The part that sticks in my mind is audio device switching on Android.) WebRTC products seem to always end up with the consumer needing to know way more of the internals than is healthy. As such I think once you are sufficiently good at using LiveKit you are less likely to pick it for your next product because you will be able to roll your own far more easily. That is unless the value you were getting from it actually was the SFU infrastructure and not the SDKs.

The OpenAI case is so point-to-point that doing WebRTC for that is, honestly, really not hard at all.

You really don’t need to know about WebRTC at all when you use LiveKit. That’s largely thanks to the SDKs abstracting away all the complexity. Having good SDKs that work across every platform with consistent APIs is more valuable than the SFU imo. There are other options for SFUs and folks like Signal have rolled their own. Try to get WebRTC running on Apple Vision Pro or tvOS and let me know if that’s no big deal.

> Try to get WebRTC running on Apple Vision Pro or tvOS and let me know if that’s no big deal.

[EDIT: I probably shouldn't mention that]. I have some experience of getting webrtc up on new platforms, and it's not as bad as all that. libwebrtc is a remarkably solid library, especially given the domain it's in.

I obviously do not share your opinion of the SDKs.

Heh, actually I'm pretty sure I've come across your X profile before. :) You're definitely in a small minority of folks with a deep(er) understanding of WebRTC.

Field CTO — hi @Sean-Der :wave:

Fractional CTO sounds like a disaster lol

My bad, he was Field CTO.

Super cool! Didn't realize OpenAI is just using LiveKit.

Does the pricing breakdown to be the same as having a OpenAI Advanced Voice socket open the whole time? It's like $9/hr!

It would be theoretically cheaper to use this without keeping the advanced voice socket open the whole time and just use the GPT4o streaming service [1] for whenever inference is needed (pay per token) and use livekits other components to do the rest (TTS, VAD etc.).

What's the trade off here?

[1]: https://platform.openai.com/docs/api-reference/streaming

Currently it does: all audio is sent to the model.

However, we are working on turn detection within the framework, so you won't have to send silence to the model when the user isn't talking. It's a fairly straight forward path to cutting down the cost by ~50%.

Working on this for an internal tool - detecting no speech has been a PITA so far. Interested to see how you go with this.

Use the voice activity detector we wrote for Home Assistant. It works very well: https://github.com/rhasspy/pymicro-vad

What if I'm watching TV and use the AI to control it ? It should only react to my voice (a problem I had that forced me to use a wake word).

currently we are using silero VAD to detect speech: https://github.com/livekit/agents/blob/main/livekit-plugins/...

it works well for voice activity; though it doesn't always detect end-of-turn correctly (humans often pause mid-sentence to think). we are working on improving this behavior.

Can I currently put a VAD module in the pipeline and only send audio when there is an active conversation? Feel like just that would solve the problem?

You dont get charged per hour with the openai realtime api, only for tokens from detected speech and response

That’s some crazy marketing for a „our library happened to support this relatively simple use case“ situation. Impressive!

By the way: The cerebras voice demo also uses LiveKit for this: https://cerebras.vercel.app/

There’s a ton of complexity under the “relatively simple use case” when you get to a global, 200M+ user scale.

80% of the times I’m experiencing choppy audio on my iPhone 15 Pro Max (18.1b) on Voice Mode (Standard and Advanced). My internet connection is FTTH and WiFi 7 state of the art router.

I wonder if this is because bugs or the crazy load livekit may be going through given the popularity in ChatGPT voice modes right now.

Doesn’t sound right. I’d love to dig into this some more. Would you mind shooting me a DM on X? @dsa

Olivier, Michelle, and Romain gave you guys a shoutout like 3 times in our DevDay recap podcast if you need more testimonial quotes :) https://www.latent.space/p/devday-2024

I had no idea! <3 Thank you for sharing this, made my weekend.

You guys are honestly the best

Is there anyone besides OpenAI working on a speech to speech model? I find it incredibly useful and it's the sole reason that I pay for their service but I do find it very limited. I'd be interested to know if any other groups are doing research on voice models.

Yes. Kyutai released an opened model called moshi : https://github.com/kyutai-labs/moshi

There's also llama-omni and a few others. None of them are even close to 4o from an LLM standpoint. But moshi is called a "foundational" model and U'm hopeful it will be enhanced. Also there's not yet support for those on most backends like llamacpp / ollama etc. So I'd say we're in a trough but we'll get there.

When I asked advanced voice mode it said that it receives input as audio and generates text as output.

It is mistaken because it has no particular insight into its own implementation. In fact the whole point is that it directly consumes and produces audio tokens with no text. That's why it's able to sing, make noises, do accents, and so on.

There’s Ultravox as well (from one of the creators of WebRTC): https://github.com/fixie-ai/ultravox

Their model builds a speech-to-speech layer into Llama. Last I checked they have the audio-in part working and they’re working on the audio-out piece.

I wonder when Azure OpenAI will get this.

I'm working on a PR now :)

This suggests that the AI "brain" receives the user input as text prompt (agent relays the speech prompt to GPT-4o) and generates audio as output (GPT-4o streams speech packets back to the agent).

But when I asked advanced voice mode it said the exact opposite. That it receives input as audio and generates text as output.

Who did you ask? ChatGPT? Not sure if you understand LLMs but its knowledge is based on the training data, it can't reason about itself, it can only hallucinate in this case, sometimes correctly, most times incorrectly.

This is also true for petty much all humans and bypassing this limitation is called enlightenment/self realization.

LLMs don't even have a self so it can never be realized. Just the ego alone exists.

No, humans can self inspect just fine

How do you know that?

Just grab a mirror.

A lot of psychologists would quibble with that...

Agreed. I am a psychologist.

Any evidence of that?

Have you seen the current US political system? Or Hawk Tua?

Both input and output are audio. This post is about bridging WebRTC audio I/O with an API that itself operates on simple TCP socket streams of raw PCM. For reliability and efficiency you want end users to connect with compressed loss-tolerant Zoom-style streams, and that goes through a middleman which relays to the model API.

Nice they have many partners on this. I see Azure as well.

There is a common consensus that the new Realtime API is not actually using the same Advanced Voice model / engine - or however it works - since at least the TTS part doesn’t seem to be as capable as the one shipped with the official OpenAI app.

Any idea on this?

Source: https://github.com/openai/openai-realtime-api-beta/issues/2

It's using the same model/engine. I don't have knowledge of the internals, but a different subsystem/set of dedicated resources though for API traffic versus first-party apps.

One thing to note is there is no separate TTS-phase here, it's happening internally within GPT-4o, in the Realtime API and Advanced Voice.

Thanks

so the WebRTC helps with the unreliable network between the mobile clients and the server side. if the application is backend only, would it make sense to use WebRTC or should I go directly to realtime api?

That was cool, but got up to $1 usage real quick

We had our playground (https://playground.livekit.io) up for a few days using our key. Def racked up a $$$$ bill!

How much is it per minute of talking?

It shakes out to around $0.15 per minute for an average conversation. If history is a guide though, this will get a lot cheaper pretty quickly.

This is cheaper than old cellular calls, inflation adjusted

50% human speaking at $0.06/minute of tokens

50% AI speaking at $0.24/minute of tokens

we (LiveKit Cloud) charge ~$0.0005/minute for each participant (in this case there would be 2)

So blended is $0.151/minute