31 lines
4.3 KiB
JSON
31 lines
4.3 KiB
JSON
{
|
|
"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.",
|
|
"pricing_note": "USD per 1M input characters, from Venice /models?type=tts on 2026-10-21",
|
|
"pricing": {"tts-kokoro": 3.5, "tts-inworld-1-5-max": 12.5, "tts-xai-v1": 18.75, "tts-gradium-v1": 47.5, "tts-chatterbox-hd": 50, "tts-orpheus": 62.5, "tts-elevenlabs-turbo-v2-5": 62.5, "tts-qwen3-1-7b": 112.5, "tts-minimax-speech-02-hd": 125, "tts-gemini-3-1-flash": 187.5},
|
|
"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},
|
|
{"id": "inworld-olivia", "family": "Inworld 1.5 Max", "model": "tts-inworld-1-5-max", "voice": "Olivia", "speed": 0.85, "format": "wav", "lead": "[whispering] ", "blurb": "Top of the quality charts. Script led with a [whispering] tag."},
|
|
{"id": "inworld-luna", "family": "Inworld 1.5 Max", "model": "tts-inworld-1-5-max", "voice": "Luna", "speed": 0.85, "format": "wav", "lead": "[whispering] "},
|
|
{"id": "inworld-luna-slow", "family": "Inworld 1.5 Max", "model": "tts-inworld-1-5-max", "voice": "Luna", "speed": 0.77, "format": "wav", "lead": "[whispering] ", "label": "Luna · slow render"},
|
|
{"id": "gemini-aoede-w", "family": "Gemini 3.1 Flash TTS", "model": "tts-gemini-3-1-flash", "voice": "Aoede", "lead": "Whisper this softly and slowly, close to the mic, like ASMR: ", "blurb": "Steered by a spoken-language direction in front of the script."},
|
|
{"id": "gemini-leda-w", "family": "Gemini 3.1 Flash TTS", "model": "tts-gemini-3-1-flash", "voice": "Leda", "lead": "Whisper this softly and slowly, close to the mic, like ASMR: "},
|
|
{"id": "minimax-calmwoman", "family": "MiniMax Speech 02 HD", "model": "tts-minimax-speech-02-hd", "voice": "CalmWoman", "speed": 0.85, "blurb": "Calm and sweet presets."},
|
|
{"id": "minimax-sweetgirl", "family": "MiniMax Speech 02 HD", "model": "tts-minimax-speech-02-hd", "voice": "SweetGirl", "speed": 0.85},
|
|
{"id": "xai-eve-w", "family": "Grok TTS", "model": "tts-xai-v1", "voice": "eve", "speed": 0.85, "lead": "<whisper>", "tail": "</whisper>", "blurb": "Wrapped in a whisper tag."},
|
|
{"id": "xai-eve-slow", "family": "Grok TTS", "model": "tts-xai-v1", "voice": "eve", "speed": 0.77, "lead": "<whisper>", "tail": "</whisper>", "label": "eve · slow render"},
|
|
{"id": "xai-ara-w", "family": "Grok TTS", "model": "tts-xai-v1", "voice": "ara", "speed": 0.85, "lead": "<whisper>", "tail": "</whisper>"}
|
|
],
|
|
"next": []
|
|
}
|