TTS Response Streaming

[allanbunch]

Hi everyone! I'm looking into implementing TTS response streaming for the node's Audio >> Create Speech method. Keep an eye on this thread for updates.

[amilanov75]

That's great news. Thank you Allan!

I played around with a direct connection, but couldn't figure out the magic combination to get my node to output anything other than a single msg with the entire file.

[allanbunch]

Curious to know if your function node direct connection attempts were similar to what's mentioned here.

[amilanov75]

Hi Allan, I am not a heavy user of GitHub. For some reason I cannot see your comment below on the comments page, so assume it is private to me. No, I have not tried what is on that page. Given some time this weekend I can look into it if you would like / if it would help? Regards Aleks

…

________________________________ From: Allan Bunch ***@***.***> Sent: Tuesday, July 2, 2024 22:51 To: allanbunch/node-red-openai-api ***@***.***> Cc: amilanov75 ***@***.***>; Comment ***@***.***> Subject: Re: [allanbunch/node-red-openai-api] TTS Response Streaming (Discussion #23) Curious to know if your function node direct connection attempts were similar to what's mentioned here<openai/openai-node#487 (comment)>. — Reply to this email directly, view it on GitHub<#23 (reply in thread)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AFVDZ3VHBWMDOBR7A5FRH33ZKMOEZAVCNFSM6AAAAABKDLAM26VHI2DSMVQWIX3LMV43SRDJONRXK43TNFXW4Q3PNVWWK3TUHM4TSNBRGUZTA>. You are receiving this because you commented.Message ID: ***@***.***>

[allanbunch]

Sounds good, Aleks. The approach suggested may be what you're looking for. Let me know if you discover otherwise.

[amilanov75]

Thank you for the prompt reply. I'll set aside some time (hopefully this weekend) to look into it and will report back.

…

________________________________ From: Allan Bunch ***@***.***> Sent: Wednesday, July 3, 2024 21:16 To: allanbunch/node-red-openai-api ***@***.***> Cc: amilanov75 ***@***.***>; Comment ***@***.***> Subject: Re: [allanbunch/node-red-openai-api] TTS Response Streaming (Discussion #23) Sounds good, Aleks. The approach suggested may be what you're looking for. Let me know if you discover otherwise. — Reply to this email directly, view it on GitHub<#23 (reply in thread)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AFVDZ3SG3FLBJU2SVUEWYADZKRLYVAVCNFSM6AAAAABKDLAM26VHI2DSMVQWIX3LMV43SRDJONRXK43TNFXW4Q3PNVWWK3TUHM4TSNJSGE3TK>. You are receiving this because you commented.Message ID: ***@***.***>

[amilanov75]

You pointed me in the exact right direction. Thank you!

I ended up getting tts to work both with a bespoke streaming text input from openai assistants API, as chunks of text, and streaming writing to the MP3 per chunk from the tts API.

It was quite a bit of work with many pitfalls and lots of design decisions on my behalf, which means there are likely other ways to get to a similar end point.

Thanks for your help!

[allanbunch]

Great to hear! Good luck with your project!