WakeNet9 custom wake word ("Hey Karen") — TTS-pipeline shortcut vs. full 20k-sample path?

[Irelove91]

I'm running WakeNet9 on an ESP32-P4 for a personal voice-controlled accessibility device (my father has a progressive neuromuscular condition; the device drives his PC's keyboard/mouse by voice so he doesn't need to touch it). wn9_jarvis_tts is working great as a placeholder — 10/10 detections at arm's length, no false triggers.

The device's actual name is "Karen," and I'd like a wn9_heykaren_tts-style model. I noticed in wakeword_list.md that a number of bundled models are TTS-generated rather than built from a collected corpus — wn9_heykira_tts3, wn9_hiandy_tts2, wn9_heyivy_tts2, wn9_sophia_tts, wn9_jarvis_tts itself, etc. "Hey Karen" looks like the same shape of request as those.

Sales (contacted directly) routed me here instead of answering, since this isn't a commercial order — it's 5 personal units. Before I start collecting the ~20,000 samples for the full custom-wake-word path in the docs: is the TTS-pipeline route open to individual/hobbyist requests at all, and if so, is there a specific esp-sr contact or process for it? Any pointer to the right person/thread would be appreciated.