156 lines
3.4 KiB
JSON
156 lines
3.4 KiB
JSON
{
|
|
"schema_version": 1,
|
|
"family": "inflect_v2",
|
|
"display_name": "Inflect Micro v2",
|
|
"description": "Compact English VITS text-to-speech models with a native GGML inference path and an external eSpeak-ng phonemizer.",
|
|
"category": "tts",
|
|
"status": "community",
|
|
"tasks": [
|
|
"tts"
|
|
],
|
|
"modes": [
|
|
"offline"
|
|
],
|
|
"languages": [
|
|
"en"
|
|
],
|
|
"runtime": {
|
|
"tags": [
|
|
"gguf"
|
|
]
|
|
},
|
|
"capabilities": {
|
|
"tts": [
|
|
"long_form"
|
|
]
|
|
},
|
|
"options": {
|
|
"request": [
|
|
{
|
|
"name": "speaking_rate",
|
|
"type": "float",
|
|
"description": "Speech speed multiplier mapped directly to Inflect speed; default 1.0.",
|
|
"required": false,
|
|
"min": 0.5,
|
|
"max": 2.0,
|
|
"default": 1.0
|
|
},
|
|
{
|
|
"name": "variation",
|
|
"type": "float",
|
|
"description": "Latent Gaussian variation; default 0.667.",
|
|
"required": false,
|
|
"min": 0.0,
|
|
"max": 1.0,
|
|
"default": 0.667
|
|
},
|
|
{
|
|
"name": "seed",
|
|
"type": "int",
|
|
"description": "Non-negative latent noise seed; default 0.",
|
|
"required": false,
|
|
"min": 0,
|
|
"default": 0
|
|
},
|
|
{
|
|
"name": "text_chunk_mode",
|
|
"type": "enum",
|
|
"description": "Long-form text chunking mode.",
|
|
"values": [
|
|
"word_budget"
|
|
],
|
|
"required": false,
|
|
"default": "word_budget"
|
|
},
|
|
{
|
|
"name": "text_chunk_size",
|
|
"type": "int",
|
|
"description": "Maximum Unicode codepoints per long-form text chunk; default 280.",
|
|
"required": false,
|
|
"min": 1,
|
|
"default": 280
|
|
}
|
|
],
|
|
"session": [
|
|
{
|
|
"name": "espeak_library_path",
|
|
"type": "path",
|
|
"description": "Optional explicit path to the eSpeak-ng shared library.",
|
|
"required": false
|
|
},
|
|
{
|
|
"name": "espeak_data_path",
|
|
"type": "path",
|
|
"description": "Optional explicit path to the directory containing espeak-ng-data.",
|
|
"required": false
|
|
}
|
|
],
|
|
"load": []
|
|
},
|
|
"package_defaults": {
|
|
"download": {
|
|
"kind": "huggingface_snapshot",
|
|
"repo": "audio-cpp/audio.cpp-gguf",
|
|
"revision": "main",
|
|
"gated": false
|
|
}
|
|
},
|
|
"packages": [
|
|
{
|
|
"id": "inflect_micro_v2_orig",
|
|
"display_name": "Inflect Micro v2 Original-Dtype GGUF",
|
|
"default": true,
|
|
"format": "gguf",
|
|
"precision": "orig",
|
|
"target_directory": "Inflect-Micro-v2-GGUF",
|
|
"files": [
|
|
"Inflect-Micro-v2-GGUF/inflect-micro-v2-orig.gguf"
|
|
],
|
|
"strip_prefix": "Inflect-Micro-v2-GGUF"
|
|
}
|
|
],
|
|
"dependencies": [],
|
|
"ui": {
|
|
"recommended_package": "inflect_micro_v2_orig",
|
|
"tags": [
|
|
"TTS",
|
|
"GGUF"
|
|
],
|
|
"docs": [
|
|
"docs/tts.md",
|
|
"docs/community_models/inflect_v2.md",
|
|
"docs/gguf.md"
|
|
]
|
|
},
|
|
"sources": [
|
|
{
|
|
"format": "gguf",
|
|
"roots": {
|
|
"model": ".",
|
|
"weights": "$gguf"
|
|
},
|
|
"files": {
|
|
"config": "model:config.json"
|
|
},
|
|
"tensors": {
|
|
"weights": {
|
|
"source": "weights:",
|
|
"prefix": "weights"
|
|
}
|
|
}
|
|
},
|
|
{
|
|
"format": "safetensors",
|
|
"roots": {
|
|
"model": "."
|
|
},
|
|
"files": {
|
|
"config": "model:config.json"
|
|
},
|
|
"tensors": {
|
|
"weights": "model:model.safetensors"
|
|
}
|
|
}
|
|
]
|
|
}
|