Question about higher sample rate support in AudioSocket (16kHz, 24kHz)

[manhcuong17072002]

Hello, first of all, thank you for developing the AudioSocket repo, which makes connecting Asterisk with external languages much more convenient compared to traditional approaches.

I have successfully tested and connected to Asterisk via AudioSocket. However, as far as I understand, AudioSocket currently only sends and receives data in 16-bit, 8kHz, mono, and transmits/receives in 320-byte (~20ms) chunks. This results in lower audio quality over the phone compared to 16kHz or 24kHz.

My question is: Does AudioSocket currently support higher sample rates such as 16kHz or 24kHz, or is there any plan for it in the future?

Thank you very much! 🙏

[Ulexus]

The original spec only indicated 16-bit signed linear, but the Asterisk team altered it after I got done, so it now works for any codec, but Asterisk never bothers to signal the codec in the header. So in that sense, it supports any codec that Asterisk supports.

The code here is pretty fixed to 16-bit 8kHz, but it should not be difficult add other dimensions. We shouldn't need to actually transcode or resample here, I think. Rather, we just need to support alternative chunking dimensions.

We allow alternative chunking sizes in some places, but not others. The main bottleneck I see is in sendAudio

[manhcuong17072002]

Thank you very much for the clear explanation 🙏.
So if I understand correctly:

• Asterisk can handle any codec, but AudioSocket here is currently fixed to 16-bit 8kHz.
• To support higher sample rates like 16kHz or 24kHz, we don’t need transcoding/resampling, but mainly need to change the chunking dimensions, especially in sendAudio.

May I ask if you could give some guidance (or example) on how to adjust sendAudio (and related parts) to properly handle 16kHz audio? That would be very helpful for us when extending the code.

Thanks again! 🙌

[Ulexus]

Yes, this is correct. I need to refamiliarise myself with this code, because I can answer the question about sendAudio. I see some code elsewhere which can handle alternative block sized, but it appears sendAudio does not.
I will try to get time over the next few days (probably the weekend) to look at this.

[manhcuong17072002]

Thank you so much for confirming.
That would be great if you could take a look when you have time. I really appreciate your support on this!
I’ll wait for your update after the weekend.

[disolaterX]

Hello guys, I was triyng to create ExternalMedia Channel via ARI and set format = alaw

and audio socket server in JS https://github.com/fonoster/fonoster/tree/main/mods/streams

but all i get is craky sound from the deepgram it's very hard to debug

[manhcuong17072002]

@disolaterX
Sorry, I haven't used ARI and the ExternalMedia channel to connect Asterisk with a third-party programming language. I did explore that approach before, but it didn't seem to work successfully. I tried recording the input audio, but it mostly resulted in noise. Since I switched to using AudioSocket, it has fulfilled all of my requirements.

[disolaterX]

exactly what i am getting it 90% noise but that not a js problem i think its a audiosocket problem at least to me it looks that way

If i create an udp media server and run external media to that it works very smoothly with 3rd party AI agents

[manhcuong17072002]

@disolaterX
Unfortunately, I’m unable to assist with that, as my testing was done using AudioSocket with Python. The only problems I experienced were related to connecting through ARI and ExternalMedia. Thank you for your understanding.

[disolaterX]

@manhcuong17072002 what are you in Python to any github repo for audiosocket implantation ?

[manhcuong17072002]

@disolaterX

I'm currently working on a company project, so unfortunately I can't share our internal repository. However, I found your repository very helpful in understanding how to work with Asterisk and stream audio to third-party applications like Golang or Python.

Here are some repos I used as references:

• audiosocket_server for Python: https://github.com/silentindark/audiosocket_server
• freetalkbot using Golang: https://github.com/felipem1210/freetalkbot

The second repo was particularly useful in helping me figure out the keys needed to connect and transmit audio data between Asterisk and external services.

If there's anything I can help with or if you're interested in discussing implementation details, feel free to reach out. I'm happy to share what I’ve learned so far!

Thanks again!

[Ulexus]

We recently merged #28 , which added support for a wide array of slin frequencies, and Asterisk now also supports these. All thanks to the efforts of @SvenKube .

I think this should resolve this issue.

[manhcuong17072002]

@Ulexus Thank you for your detailed response and for enabling higher input/output format support for enhanced audio fidelity. I’d like to clarify whether, when connected via a telecom carrier, these formats are constrained by the carrier’s transmission characteristics or can be configured independently within AudioSocket. Additionally, is it necessary to upgrade Asterisk to a newer release to ensure compatibility with higher sample rates like 16 kHz or 24 kHz? Thank you.

[Ulexus]

Previously to Sven's changes, Asterisk completely ignored the AudioSocket specification of codec. Whatever codec you set for the channel is what the Asterisk side of AudioSocket sent (and in turn, expected to receive). In that sense, you could use whatever codec Asterisk supported; you just needed to manually coordinate between the application side and the Asterisk side.

After the changes, the protocol header is set correctly, so the application side can adjust the codec accordingly.

As to whether the carrier supports the higher bit-rate codec: standard PSTN is still 8kHz. However, many mobile networks now support higher bitrates, and of course, if you never touch the legacy PSTN, you are not constrained thereby. Some carriers therefore do support the higher bitrate codecs, but that will vary on a carrier-by-carrier basis, and the specific set of codecs they may support will also do so.

[manhcuong17072002]

@Ulexus Thank you very much for the clear and detailed explanation.