Default Orpheus TTS voice to tara on Together provider

[nbroad1881]

Summary

• Default the Together text-to-speech voice to tara when the model id contains orpheus (e.g. canopylabs/orpheus-3b-0.1-ft).
• Keeps the existing Kokoro default (af_alloy) and only applies defaults when no parameters.voice is provided.

Test plan

• Call textToSpeech with provider: "together" and model: "canopylabs/orpheus-3b-0.1-ft" without a voice parameter and confirm audio is generated.
• Confirm an explicit parameters.voice still overrides the default.
• Confirm Kokoro models still default to af_alloy.

Made with Cursor

---

Note

Low Risk
Small provider payload tweak for optional voice defaults; no auth, data, or API surface changes beyond TTS request shaping.

Overview
Together text-to-speech now picks a default voice when callers omit parameters.voice: Kokoro models still get af_alloy, and model ids containing orpheus get tara. Other models still send no default voice unless the user sets one.

Explicit parameters.voice continues to override these defaults.

^{Reviewed by Cursor Bugbot for commit 953e979. Bugbot is set up for automated code reviews on this repo. Configure here.}