Open ai Webrtc on ADF (AUD-6538)

[kungfupizza]

Is there a feature backlog with OPEN AI WebRTC in the esp-adf maybe as a pipeline option.

[TempoTian]

We have commited openai solution into esp-webrtc-solution repo: https://github.com/espressif/esp-webrtc-solution/tree/main/solutions/openai_demo. It is a complete solution if use pipeline, still need build it from scratch and may takes more efforts.
And we are working on to develop esp_capture(Finished) and esp_render (Ongoing) which use ESP-GMF (similar with pipeline but use less resources), and they will replace esp-webrtc-solution esp_capture and av_render modules. Both the 2 new developed modules will provide flexible configuration for pipeline building and customization.
So we are not decided to add pipeine support on ADF for openAI. Do you have some special concerns? Please let me know.

[kungfupizza]

Hi,
Thank you for the response. I have looked at your openai_demo repo but does it use the ADF pipeline?

I have a personal project that uses pipeline for playing audio from SD card but when I try to use WebRTC in parallel to that, there is a lot of garbage audio after 3-4 seconds so I was thinking to maybe use WebRTC solution on a pipeline to have a common development.

I can possibly wait for the GMF development and then convert my existing code to GMF but I think that'll be months before it is out.

[TempoTian]

I think you can attach your test file, if possible can provide the test code, I will help you to check the possible issue point. You can raise a issue on esp-webrtc-solution branch. av_render is part of a player project and has been tested for variaties of file format, I think there maybe some configuration errors.