| 1 | import 'package:flutter/material.dart'; |
| 2 | |
| 3 | /// Available KittenTTS model variants downloadable from HuggingFace. |
| 4 | /// |
| 5 | /// Each variant has a different parameter count and quality trade-off. |
| 6 | enum KittenTtsModelVariant { |
| 7 | /// 15M parameters, ~56 MB total, high quality. |
| 8 | nano, |
| 9 | |
| 10 | /// 15M parameters, ~25 MB total, int8 quantized. Best for most devices. |
| 11 | nanoInt8, |
| 12 | |
| 13 | /// 40M parameters, ~41 MB total, higher quality. |
| 14 | micro, |
| 15 | |
| 16 | /// 80M parameters, ~80 MB total, best quality. |
| 17 | mini; |
| 18 | |
| 19 | String get displayName { |
| 20 | switch (this) { |
| 21 | case KittenTtsModelVariant.nano: |
| 22 | return 'Kitten TTS Nano'; |
| 23 | case KittenTtsModelVariant.nanoInt8: |
| 24 | return 'Kitten TTS Nano (Int8)'; |
| 25 | case KittenTtsModelVariant.micro: |
| 26 | return 'Kitten TTS Micro'; |
| 27 | case KittenTtsModelVariant.mini: |
| 28 | return 'Kitten TTS Mini'; |
| 29 | } |
| 30 | } |
| 31 | |
| 32 | /// Short label shown in badges (e.g. "15M", "80M"). |
| 33 | String get parameterLabel { |
| 34 | switch (this) { |
| 35 | case KittenTtsModelVariant.nano: |
| 36 | return '15M'; |
| 37 | case KittenTtsModelVariant.nanoInt8: |
| 38 | return '15M'; |
| 39 | case KittenTtsModelVariant.micro: |
| 40 | return '40M'; |
| 41 | case KittenTtsModelVariant.mini: |
| 42 | return '80M'; |
| 43 | } |
| 44 | } |
| 45 | |
| 46 | /// Approximate total size in bytes for all 3 files combined. |
| 47 | int get totalSizeBytes { |
| 48 | switch (this) { |
| 49 | case KittenTtsModelVariant.nano: |
| 50 | return 56 * 1024 * 1024; |
| 51 | case KittenTtsModelVariant.nanoInt8: |
| 52 | return 25 * 1024 * 1024; |
| 53 | case KittenTtsModelVariant.micro: |
| 54 | return 41 * 1024 * 1024; |
| 55 | case KittenTtsModelVariant.mini: |
| 56 | return 80 * 1024 * 1024; |
| 57 | } |
| 58 | } |
| 59 | |
| 60 | String get description { |
| 61 | switch (this) { |
| 62 | case KittenTtsModelVariant.nano: |
| 63 | return 'Balanced quality and size. Good for most devices.'; |
| 64 | case KittenTtsModelVariant.nanoInt8: |
| 65 | return 'Smallest model with excellent quality. Recommended for most devices.'; |
| 66 | case KittenTtsModelVariant.micro: |
| 67 | return 'Higher quality with more parameters. Requires more RAM.'; |
| 68 | case KittenTtsModelVariant.mini: |
| 69 | return 'Highest quality, largest size. Best for high-end devices.'; |
| 70 | } |
| 71 | } |
| 72 | |
| 73 | /// Whether this variant is recommended as the default. |
| 74 | bool get isRecommended => this == KittenTtsModelVariant.nanoInt8; |
| 75 | |
| 76 | /// HuggingFace repository ID for this variant. |
| 77 | String get huggingFaceRepoId { |
| 78 | switch (this) { |
| 79 | case KittenTtsModelVariant.nano: |
| 80 | return 'palshub/kitten-tts-nano-0.8-fp32'; |
| 81 | case KittenTtsModelVariant.nanoInt8: |
| 82 | return 'palshub/kitten-tts-nano-0.8-int8'; |
| 83 | case KittenTtsModelVariant.micro: |
| 84 | return 'palshub/kitten-tts-micro-0.8'; |
| 85 | case KittenTtsModelVariant.mini: |
| 86 | return 'palshub/kitten-tts-mini-0.8'; |
| 87 | } |
| 88 | } |
| 89 | |
| 90 | /// The ONNX model file name inside the HF repo. |
| 91 | String get sourceModelFileName { |
| 92 | switch (this) { |
| 93 | case KittenTtsModelVariant.nano: |
| 94 | return 'kitten_tts_nano_v0_8.onnx'; |
| 95 | case KittenTtsModelVariant.nanoInt8: |
| 96 | return 'kitten_tts_nano_v0_8.onnx'; |
| 97 | case KittenTtsModelVariant.micro: |
| 98 | return 'kitten_tts_micro_v0_8.onnx'; |
| 99 | case KittenTtsModelVariant.mini: |
| 100 | return 'kitten_tts_mini_v0_8.onnx'; |
| 101 | } |
| 102 | } |
| 103 | |
| 104 | /// sherpa-onnx bundle directory for this variant. |
| 105 | String get bundleDirName { |
| 106 | switch (this) { |
| 107 | case KittenTtsModelVariant.nano: |
| 108 | return 'kitten-nano-en-v0_1-fp16'; |
| 109 | case KittenTtsModelVariant.nanoInt8: |
| 110 | return 'kitten-nano-en-v0_1-fp16'; |
| 111 | case KittenTtsModelVariant.micro: |
| 112 | return 'kitten-nano-en-v0_2-fp16'; |
| 113 | case KittenTtsModelVariant.mini: |
| 114 | return 'kitten-mini-en-v0_1-fp16'; |
| 115 | } |
| 116 | } |
| 117 | |
| 118 | String get tarballUrl { |
| 119 | switch (this) { |
| 120 | case KittenTtsModelVariant.nano: |
| 121 | case KittenTtsModelVariant.nanoInt8: |
| 122 | return 'https://github.com/k2-fsa/sherpa-onnx/releases/download/tts-models/kitten-nano-en-v0_1-fp16.tar.bz2'; |
| 123 | case KittenTtsModelVariant.micro: |
| 124 | return 'https://github.com/k2-fsa/sherpa-onnx/releases/download/tts-models/kitten-nano-en-v0_2-fp16.tar.bz2'; |
| 125 | case KittenTtsModelVariant.mini: |
| 126 | return 'https://github.com/k2-fsa/sherpa-onnx/releases/download/tts-models/kitten-mini-en-v0_1-fp16.tar.bz2'; |
| 127 | } |
| 128 | } |
| 129 | |
| 130 | String get modelFileName => 'model.fp16.onnx'; |
| 131 | String get voicesFileName => 'voices.bin'; |
| 132 | String get tokensFileName => 'tokens.txt'; |
| 133 | String get dataDirName => 'espeak-ng-data'; |
| 134 | |
| 135 | /// The 3 files required for this variant. |
| 136 | List<KittenTtsModelFile> get files { |
| 137 | return [ |
| 138 | KittenTtsModelFile( |
| 139 | fileName: 'config.json', |
| 140 | downloadUrl: |
| 141 | 'https://huggingface.co/$huggingFaceRepoId/resolve/main/config.json', |
| 142 | sizeBytes: 2 * 1024, |
| 143 | ), |
| 144 | KittenTtsModelFile( |
| 145 | fileName: sourceModelFileName, |
| 146 | downloadUrl: |
| 147 | 'https://huggingface.co/$huggingFaceRepoId/resolve/main/$sourceModelFileName', |
| 148 | sizeBytes: _modelFileSizeBytes, |
| 149 | ), |
| 150 | KittenTtsModelFile( |
| 151 | fileName: 'voices.npz', |
| 152 | downloadUrl: |
| 153 | 'https://huggingface.co/$huggingFaceRepoId/resolve/main/voices.npz', |
| 154 | sizeBytes: 3 * 1024 * 1024, |
| 155 | ), |
| 156 | ]; |
| 157 | } |
| 158 | |
| 159 | int get _modelFileSizeBytes { |
| 160 | switch (this) { |
| 161 | case KittenTtsModelVariant.nano: |
| 162 | return 52 * 1024 * 1024; |
| 163 | case KittenTtsModelVariant.nanoInt8: |
| 164 | return 21 * 1024 * 1024; |
| 165 | case KittenTtsModelVariant.micro: |
| 166 | return 37 * 1024 * 1024; |
| 167 | case KittenTtsModelVariant.mini: |
| 168 | return 76 * 1024 * 1024; |
| 169 | } |
| 170 | } |
| 171 | |
| 172 | /// Storage directory name for this variant. |
| 173 | String get dirName => name; |
| 174 | } |
| 175 | |
| 176 | /// Metadata for a single file within a KittenTTS model variant. |
| 177 | @immutable |
| 178 | class KittenTtsModelFile { |
| 179 | final String fileName; |
| 180 | final String downloadUrl; |
| 181 | final int sizeBytes; |
| 182 | |
| 183 | const KittenTtsModelFile({ |
| 184 | required this.fileName, |
| 185 | required this.downloadUrl, |
| 186 | required this.sizeBytes, |
| 187 | }); |
| 188 | } |
| 189 | |
| 190 | /// A fully-specified KittenTTS model with its variant and files. |
| 191 | @immutable |
| 192 | class KittenTtsModel { |
| 193 | final KittenTtsModelVariant variant; |
| 194 | final List<KittenTtsModelFile> files; |
| 195 | final int totalSizeBytes; |
| 196 | |
| 197 | const KittenTtsModel({ |
| 198 | required this.variant, |
| 199 | required this.files, |
| 200 | required this.totalSizeBytes, |
| 201 | }); |
| 202 | |
| 203 | static final List<KittenTtsModel> allModels = [ |
| 204 | KittenTtsModel( |
| 205 | variant: KittenTtsModelVariant.nano, |
| 206 | files: KittenTtsModelVariant.nano.files, |
| 207 | totalSizeBytes: KittenTtsModelVariant.nano.totalSizeBytes, |
| 208 | ), |
| 209 | KittenTtsModel( |
| 210 | variant: KittenTtsModelVariant.nanoInt8, |
| 211 | files: KittenTtsModelVariant.nanoInt8.files, |
| 212 | totalSizeBytes: KittenTtsModelVariant.nanoInt8.totalSizeBytes, |
| 213 | ), |
| 214 | KittenTtsModel( |
| 215 | variant: KittenTtsModelVariant.micro, |
| 216 | files: KittenTtsModelVariant.micro.files, |
| 217 | totalSizeBytes: KittenTtsModelVariant.micro.totalSizeBytes, |
| 218 | ), |
| 219 | KittenTtsModel( |
| 220 | variant: KittenTtsModelVariant.mini, |
| 221 | files: KittenTtsModelVariant.mini.files, |
| 222 | totalSizeBytes: KittenTtsModelVariant.mini.totalSizeBytes, |
| 223 | ), |
| 224 | ]; |
| 225 | |
| 226 | String get displayName => variant.displayName; |
| 227 | |
| 228 | String get parameterLabel => variant.parameterLabel; |
| 229 | |
| 230 | String get description => variant.description; |
| 231 | |
| 232 | bool get isRecommended => variant.isRecommended; |
| 233 | } |
| 234 | |
| 235 | /// Progress for downloading a single file of a model variant. |
| 236 | @immutable |
| 237 | class KittenTtsFileProgress { |
| 238 | final String fileName; |
| 239 | final KittenTtsModelVariant variant; |
| 240 | final int receivedBytes; |
| 241 | final int totalBytes; |
| 242 | final bool isComplete; |
| 243 | |
| 244 | const KittenTtsFileProgress({ |
| 245 | required this.fileName, |
| 246 | required this.variant, |
| 247 | this.receivedBytes = 0, |
| 248 | this.totalBytes = 0, |
| 249 | this.isComplete = false, |
| 250 | }); |
| 251 | |
| 252 | double get fraction => |
| 253 | totalBytes > 0 ? (receivedBytes / totalBytes).clamp(0.0, 1.0) : 0.0; |
| 254 | } |