diff --git a/meta/tts-voice.md b/meta/tts-voice.md
new file mode 100644
index 0000000..ce88fae
--- /dev/null
+++ b/meta/tts-voice.md
@@ -0,0 +1,20 @@
+# TTS & Voice
+
+## Endpoints
+- **STT (Whisper):** http://192.168.1.218:39000/v1 — model: `Systran/faster-whisper-base` 🔲
+- **Piper TTS:** http://192.168.1.218:39001/v1 🔲
+
+## Hermes TTS Config (from config.yaml)
+- **Active provider:** edge
+- **Edge voice:** en-US-AriaNeural
+- **Piper voice (if switched):** en_US-lessac-medium
+
+## STT Config
+- **Provider:** local
+- **Model:** base
+- **Enabled:** true
+
+## Notes
+- Both STT and TTS endpoints at 192.168.1.218 — not yet verified by this agent
+- Discord voice FX disabled in current config
+- **Voice interaction requirement (confirmed 2026-07-31):** generated responses must be human-friendly, clear, and conversational. The human participant will guide and ask follow-up questions, so prefer spoken-language turns over dense written-style answers.