Release Notes: v0.8.0¶
Release Date: 2026-10-28
Highlights¶
- Phone-call speech: the OmniVoice TTS provider returns 8 kHz mu-law or A-law audio, so OpenAI voices can speak on Twilio, Telnyx, and other phone calls, and can be swapped with other OmniVoice providers without changing the synthesis config
Added¶
OutputFormat: "mulaw"(or"ulaw") and"alaw"return 8 kHz audio unlessSampleRateasks for another rate. OpenAI produces raw speech only as 24 kHz PCM, so the provider converts itOutputFormat: "pcm"with aSampleRateother than 24000 is resampled to that rate- Format names with the rate appended, such as
ulaw_8000andpcm_16000, are accepted, so a config written for another provider works unchanged SynthesizeStreamconverts each chunk the same way. A chunk that ends partway through a sample is held until the next one, so streamed audio matches the whole-clip conversion byte for bytePCMSampleRateconstant for the rate of OpenAI's raw speech
Fixed¶
SynthesisResult.SampleRateis set for raw formats; it was 0
Dependencies¶
modernc.org/sqlitefrom v1.60.1 to v1.61.0- Indirect dependencies updated
Installation¶
go get github.com/plexusone/[email protected]