From 5e5b3b9e665fe411b855760988b718dab235e7d9 Mon Sep 17 00:00:00 2001
From: teahubxyz <boring+teahub@agentmail.to>
Date: Wed, 21 Oct 2026 18:23:46 +0000
Subject: [PATCH] asmr: voice bake-off config

---
 asmr/voices.json | 23 +++++++++++++++++++++++
 1 file changed, 23 insertions(+)
 create mode 100644 asmr/voices.json

diff --git a/asmr/voices.json b/asmr/voices.json
new file mode 100644
index 0000000..73decec
--- /dev/null
+++ b/asmr/voices.json
@@ -0,0 +1,23 @@
+{
+  "note": "One script, many voices. tools/asmr.py renders each entry through inference.yolopreneur.lol /v1/speech into asmr/audio/<id>.mp3. Edit and push to re-render (existing files are kept; delete one to redo it).",
+  "script": "Hey... you're still up. ... Come here, closer. ... Just breathe with me for a second. ... In... ... and out. ... Can you hear that? The rain's started. ... I'll stay right here... until you fall asleep. ... Goodnight.",
+  "style": "Very soft ASMR whisper, breathy, intimate, close to the microphone, slow and gentle, long pauses.",
+  "voices": [
+    {"id": "kokoro-nicole", "family": "Kokoro", "model": "tts-kokoro", "voice": "af_nicole", "speed": 0.8, "blurb": "Baseline. What gandora used. Nicole is Kokoro's breathiest voice."},
+    {"id": "kokoro-sky", "family": "Kokoro", "model": "tts-kokoro", "voice": "af_sky", "speed": 0.8},
+    {"id": "orpheus-tara", "family": "Orpheus", "model": "tts-orpheus", "voice": "tara", "speed": 0.85, "format": "wav", "tags": true, "blurb": "Emotive tags (<sigh>) in the text."},
+    {"id": "orpheus-mia", "family": "Orpheus", "model": "tts-orpheus", "voice": "mia", "speed": 0.85, "format": "wav", "tags": true},
+    {"id": "qwen3-serena", "family": "Qwen3 TTS", "model": "tts-qwen3-1-7b", "voice": "Serena", "style": true, "blurb": "Takes a written style direction: the whisper prompt above."},
+    {"id": "qwen3-vivian", "family": "Qwen3 TTS", "model": "tts-qwen3-1-7b", "voice": "Vivian", "style": true},
+    {"id": "chatterbox-aurora", "family": "Chatterbox HD", "model": "tts-chatterbox-hd", "voice": "Aurora", "speed": 0.85, "format": "wav", "blurb": "Expressive open model."},
+    {"id": "chatterbox-siobhan", "family": "Chatterbox HD", "model": "tts-chatterbox-hd", "voice": "Siobhan", "speed": 0.85, "format": "wav"},
+    {"id": "elevenlabs-lily", "family": "ElevenLabs Turbo 2.5", "model": "tts-elevenlabs-turbo-v2-5", "voice": "Lily", "speed": 0.85, "blurb": "Polished, but no whisper tags in this generation."},
+    {"id": "elevenlabs-charlotte", "family": "ElevenLabs Turbo 2.5", "model": "tts-elevenlabs-turbo-v2-5", "voice": "Charlotte", "speed": 0.85}
+  ],
+  "next": [
+    {"family": "Inworld 1.5 Max", "model": "tts-inworld-1-5-max", "why": "top of the quality charts; whisper-style markup"},
+    {"family": "Gemini 3.1 Flash TTS", "model": "tts-gemini-3-1-flash", "why": "steered by natural-language direction"},
+    {"family": "MiniMax Speech 02 HD", "model": "tts-minimax-speech-02-hd", "why": "emotion controls; CalmWoman, SweetGirl"},
+    {"family": "Grok TTS", "model": "tts-xai-v1", "why": "untested"}
+  ]
+}