polly.d.ts 28.3 KB
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510
import {Request} from '../lib/request';
import {Response} from '../lib/response';
import {AWSError} from '../lib/error';
import {PollyCustomizations} from '../lib/services/polly';
import {ServiceConfigurationOptions} from '../lib/service';
import {ConfigBase as Config} from '../lib/config';
import {Presigner as presigner} from '../lib/polly/presigner';
import {Readable} from 'stream';
interface Blob {}
declare class Polly extends PollyCustomizations {
  /**
   * Constructs a service object. This object has one method for each API operation.
   */
  constructor(options?: Polly.Types.ClientConfiguration)
  config: Config & Polly.Types.ClientConfiguration;
  /**
   * Deletes the specified pronunciation lexicon stored in an AWS Region. A lexicon which has been deleted is not available for speech synthesis, nor is it possible to retrieve it using either the GetLexicon or ListLexicon APIs. For more information, see Managing Lexicons.
   */
  deleteLexicon(params: Polly.Types.DeleteLexiconInput, callback?: (err: AWSError, data: Polly.Types.DeleteLexiconOutput) => void): Request<Polly.Types.DeleteLexiconOutput, AWSError>;
  /**
   * Deletes the specified pronunciation lexicon stored in an AWS Region. A lexicon which has been deleted is not available for speech synthesis, nor is it possible to retrieve it using either the GetLexicon or ListLexicon APIs. For more information, see Managing Lexicons.
   */
  deleteLexicon(callback?: (err: AWSError, data: Polly.Types.DeleteLexiconOutput) => void): Request<Polly.Types.DeleteLexiconOutput, AWSError>;
  /**
   * Returns the list of voices that are available for use when requesting speech synthesis. Each voice speaks a specified language, is either male or female, and is identified by an ID, which is the ASCII version of the voice name.  When synthesizing speech ( SynthesizeSpeech ), you provide the voice ID for the voice you want from the list of voices returned by DescribeVoices. For example, you want your news reader application to read news in a specific language, but giving a user the option to choose the voice. Using the DescribeVoices operation you can provide the user with a list of available voices to select from.  You can optionally specify a language code to filter the available voices. For example, if you specify en-US, the operation returns a list of all available US English voices.  This operation requires permissions to perform the polly:DescribeVoices action.
   */
  describeVoices(params: Polly.Types.DescribeVoicesInput, callback?: (err: AWSError, data: Polly.Types.DescribeVoicesOutput) => void): Request<Polly.Types.DescribeVoicesOutput, AWSError>;
  /**
   * Returns the list of voices that are available for use when requesting speech synthesis. Each voice speaks a specified language, is either male or female, and is identified by an ID, which is the ASCII version of the voice name.  When synthesizing speech ( SynthesizeSpeech ), you provide the voice ID for the voice you want from the list of voices returned by DescribeVoices. For example, you want your news reader application to read news in a specific language, but giving a user the option to choose the voice. Using the DescribeVoices operation you can provide the user with a list of available voices to select from.  You can optionally specify a language code to filter the available voices. For example, if you specify en-US, the operation returns a list of all available US English voices.  This operation requires permissions to perform the polly:DescribeVoices action.
   */
  describeVoices(callback?: (err: AWSError, data: Polly.Types.DescribeVoicesOutput) => void): Request<Polly.Types.DescribeVoicesOutput, AWSError>;
  /**
   * Returns the content of the specified pronunciation lexicon stored in an AWS Region. For more information, see Managing Lexicons.
   */
  getLexicon(params: Polly.Types.GetLexiconInput, callback?: (err: AWSError, data: Polly.Types.GetLexiconOutput) => void): Request<Polly.Types.GetLexiconOutput, AWSError>;
  /**
   * Returns the content of the specified pronunciation lexicon stored in an AWS Region. For more information, see Managing Lexicons.
   */
  getLexicon(callback?: (err: AWSError, data: Polly.Types.GetLexiconOutput) => void): Request<Polly.Types.GetLexiconOutput, AWSError>;
  /**
   * Retrieves a specific SpeechSynthesisTask object based on its TaskID. This object contains information about the given speech synthesis task, including the status of the task, and a link to the S3 bucket containing the output of the task.
   */
  getSpeechSynthesisTask(params: Polly.Types.GetSpeechSynthesisTaskInput, callback?: (err: AWSError, data: Polly.Types.GetSpeechSynthesisTaskOutput) => void): Request<Polly.Types.GetSpeechSynthesisTaskOutput, AWSError>;
  /**
   * Retrieves a specific SpeechSynthesisTask object based on its TaskID. This object contains information about the given speech synthesis task, including the status of the task, and a link to the S3 bucket containing the output of the task.
   */
  getSpeechSynthesisTask(callback?: (err: AWSError, data: Polly.Types.GetSpeechSynthesisTaskOutput) => void): Request<Polly.Types.GetSpeechSynthesisTaskOutput, AWSError>;
  /**
   * Returns a list of pronunciation lexicons stored in an AWS Region. For more information, see Managing Lexicons.
   */
  listLexicons(params: Polly.Types.ListLexiconsInput, callback?: (err: AWSError, data: Polly.Types.ListLexiconsOutput) => void): Request<Polly.Types.ListLexiconsOutput, AWSError>;
  /**
   * Returns a list of pronunciation lexicons stored in an AWS Region. For more information, see Managing Lexicons.
   */
  listLexicons(callback?: (err: AWSError, data: Polly.Types.ListLexiconsOutput) => void): Request<Polly.Types.ListLexiconsOutput, AWSError>;
  /**
   * Returns a list of SpeechSynthesisTask objects ordered by their creation date. This operation can filter the tasks by their status, for example, allowing users to list only tasks that are completed.
   */
  listSpeechSynthesisTasks(params: Polly.Types.ListSpeechSynthesisTasksInput, callback?: (err: AWSError, data: Polly.Types.ListSpeechSynthesisTasksOutput) => void): Request<Polly.Types.ListSpeechSynthesisTasksOutput, AWSError>;
  /**
   * Returns a list of SpeechSynthesisTask objects ordered by their creation date. This operation can filter the tasks by their status, for example, allowing users to list only tasks that are completed.
   */
  listSpeechSynthesisTasks(callback?: (err: AWSError, data: Polly.Types.ListSpeechSynthesisTasksOutput) => void): Request<Polly.Types.ListSpeechSynthesisTasksOutput, AWSError>;
  /**
   * Stores a pronunciation lexicon in an AWS Region. If a lexicon with the same name already exists in the region, it is overwritten by the new lexicon. Lexicon operations have eventual consistency, therefore, it might take some time before the lexicon is available to the SynthesizeSpeech operation. For more information, see Managing Lexicons.
   */
  putLexicon(params: Polly.Types.PutLexiconInput, callback?: (err: AWSError, data: Polly.Types.PutLexiconOutput) => void): Request<Polly.Types.PutLexiconOutput, AWSError>;
  /**
   * Stores a pronunciation lexicon in an AWS Region. If a lexicon with the same name already exists in the region, it is overwritten by the new lexicon. Lexicon operations have eventual consistency, therefore, it might take some time before the lexicon is available to the SynthesizeSpeech operation. For more information, see Managing Lexicons.
   */
  putLexicon(callback?: (err: AWSError, data: Polly.Types.PutLexiconOutput) => void): Request<Polly.Types.PutLexiconOutput, AWSError>;
  /**
   * Allows the creation of an asynchronous synthesis task, by starting a new SpeechSynthesisTask. This operation requires all the standard information needed for speech synthesis, plus the name of an Amazon S3 bucket for the service to store the output of the synthesis task and two optional parameters (OutputS3KeyPrefix and SnsTopicArn). Once the synthesis task is created, this operation will return a SpeechSynthesisTask object, which will include an identifier of this task as well as the current status.
   */
  startSpeechSynthesisTask(params: Polly.Types.StartSpeechSynthesisTaskInput, callback?: (err: AWSError, data: Polly.Types.StartSpeechSynthesisTaskOutput) => void): Request<Polly.Types.StartSpeechSynthesisTaskOutput, AWSError>;
  /**
   * Allows the creation of an asynchronous synthesis task, by starting a new SpeechSynthesisTask. This operation requires all the standard information needed for speech synthesis, plus the name of an Amazon S3 bucket for the service to store the output of the synthesis task and two optional parameters (OutputS3KeyPrefix and SnsTopicArn). Once the synthesis task is created, this operation will return a SpeechSynthesisTask object, which will include an identifier of this task as well as the current status.
   */
  startSpeechSynthesisTask(callback?: (err: AWSError, data: Polly.Types.StartSpeechSynthesisTaskOutput) => void): Request<Polly.Types.StartSpeechSynthesisTaskOutput, AWSError>;
  /**
   * Synthesizes UTF-8 input, plain text or SSML, to a stream of bytes. SSML input must be valid, well-formed SSML. Some alphabets might not be available with all the voices (for example, Cyrillic might not be read at all by English voices) unless phoneme mapping is used. For more information, see How it Works.
   */
  synthesizeSpeech(params: Polly.Types.SynthesizeSpeechInput, callback?: (err: AWSError, data: Polly.Types.SynthesizeSpeechOutput) => void): Request<Polly.Types.SynthesizeSpeechOutput, AWSError>;
  /**
   * Synthesizes UTF-8 input, plain text or SSML, to a stream of bytes. SSML input must be valid, well-formed SSML. Some alphabets might not be available with all the voices (for example, Cyrillic might not be read at all by English voices) unless phoneme mapping is used. For more information, see How it Works.
   */
  synthesizeSpeech(callback?: (err: AWSError, data: Polly.Types.SynthesizeSpeechOutput) => void): Request<Polly.Types.SynthesizeSpeechOutput, AWSError>;
}
declare namespace Polly {
  export import Presigner = presigner;
}
declare namespace Polly {
  export type Alphabet = string;
  export type AudioStream = Buffer|Uint8Array|Blob|string|Readable;
  export type ContentType = string;
  export type DateTime = Date;
  export interface DeleteLexiconInput {
    /**
     * The name of the lexicon to delete. Must be an existing lexicon in the region.
     */
    Name: LexiconName;
  }
  export interface DeleteLexiconOutput {
  }
  export interface DescribeVoicesInput {
    /**
     * Specifies the engine (standard or neural) used by Amazon Polly when processing input text for speech synthesis. 
     */
    Engine?: Engine;
    /**
     *  The language identification tag (ISO 639 code for the language name-ISO 3166 country code) for filtering the list of voices returned. If you don't specify this optional parameter, all available voices are returned. 
     */
    LanguageCode?: LanguageCode;
    /**
     * Boolean value indicating whether to return any bilingual voices that use the specified language as an additional language. For instance, if you request all languages that use US English (es-US), and there is an Italian voice that speaks both Italian (it-IT) and US English, that voice will be included if you specify yes but not if you specify no.
     */
    IncludeAdditionalLanguageCodes?: IncludeAdditionalLanguageCodes;
    /**
     * An opaque pagination token returned from the previous DescribeVoices operation. If present, this indicates where to continue the listing.
     */
    NextToken?: NextToken;
  }
  export interface DescribeVoicesOutput {
    /**
     * A list of voices with their properties.
     */
    Voices?: VoiceList;
    /**
     * The pagination token to use in the next request to continue the listing of voices. NextToken is returned only if the response is truncated.
     */
    NextToken?: NextToken;
  }
  export type Engine = "standard"|"neural"|string;
  export type EngineList = Engine[];
  export type Gender = "Female"|"Male"|string;
  export interface GetLexiconInput {
    /**
     * Name of the lexicon.
     */
    Name: LexiconName;
  }
  export interface GetLexiconOutput {
    /**
     * Lexicon object that provides name and the string content of the lexicon. 
     */
    Lexicon?: Lexicon;
    /**
     * Metadata of the lexicon, including phonetic alphabetic used, language code, lexicon ARN, number of lexemes defined in the lexicon, and size of lexicon in bytes.
     */
    LexiconAttributes?: LexiconAttributes;
  }
  export interface GetSpeechSynthesisTaskInput {
    /**
     * The Amazon Polly generated identifier for a speech synthesis task.
     */
    TaskId: TaskId;
  }
  export interface GetSpeechSynthesisTaskOutput {
    /**
     * SynthesisTask object that provides information from the requested task, including output format, creation time, task status, and so on.
     */
    SynthesisTask?: SynthesisTask;
  }
  export type IncludeAdditionalLanguageCodes = boolean;
  export type LanguageCode = "arb"|"cmn-CN"|"cy-GB"|"da-DK"|"de-DE"|"en-AU"|"en-GB"|"en-GB-WLS"|"en-IN"|"en-US"|"es-ES"|"es-MX"|"es-US"|"fr-CA"|"fr-FR"|"is-IS"|"it-IT"|"ja-JP"|"hi-IN"|"ko-KR"|"nb-NO"|"nl-NL"|"pl-PL"|"pt-BR"|"pt-PT"|"ro-RO"|"ru-RU"|"sv-SE"|"tr-TR"|string;
  export type LanguageCodeList = LanguageCode[];
  export type LanguageName = string;
  export type LastModified = Date;
  export type LexemesCount = number;
  export interface Lexicon {
    /**
     * Lexicon content in string format. The content of a lexicon must be in PLS format.
     */
    Content?: LexiconContent;
    /**
     * Name of the lexicon.
     */
    Name?: LexiconName;
  }
  export type LexiconArn = string;
  export interface LexiconAttributes {
    /**
     * Phonetic alphabet used in the lexicon. Valid values are ipa and x-sampa.
     */
    Alphabet?: Alphabet;
    /**
     * Language code that the lexicon applies to. A lexicon with a language code such as "en" would be applied to all English languages (en-GB, en-US, en-AUS, en-WLS, and so on.
     */
    LanguageCode?: LanguageCode;
    /**
     * Date lexicon was last modified (a timestamp value).
     */
    LastModified?: LastModified;
    /**
     * Amazon Resource Name (ARN) of the lexicon.
     */
    LexiconArn?: LexiconArn;
    /**
     * Number of lexemes in the lexicon.
     */
    LexemesCount?: LexemesCount;
    /**
     * Total size of the lexicon, in characters.
     */
    Size?: Size;
  }
  export type LexiconContent = string;
  export interface LexiconDescription {
    /**
     * Name of the lexicon.
     */
    Name?: LexiconName;
    /**
     * Provides lexicon metadata.
     */
    Attributes?: LexiconAttributes;
  }
  export type LexiconDescriptionList = LexiconDescription[];
  export type LexiconName = string;
  export type LexiconNameList = LexiconName[];
  export interface ListLexiconsInput {
    /**
     * An opaque pagination token returned from previous ListLexicons operation. If present, indicates where to continue the list of lexicons.
     */
    NextToken?: NextToken;
  }
  export interface ListLexiconsOutput {
    /**
     * A list of lexicon names and attributes.
     */
    Lexicons?: LexiconDescriptionList;
    /**
     * The pagination token to use in the next request to continue the listing of lexicons. NextToken is returned only if the response is truncated.
     */
    NextToken?: NextToken;
  }
  export interface ListSpeechSynthesisTasksInput {
    /**
     * Maximum number of speech synthesis tasks returned in a List operation.
     */
    MaxResults?: MaxResults;
    /**
     * The pagination token to use in the next request to continue the listing of speech synthesis tasks. 
     */
    NextToken?: NextToken;
    /**
     * Status of the speech synthesis tasks returned in a List operation
     */
    Status?: TaskStatus;
  }
  export interface ListSpeechSynthesisTasksOutput {
    /**
     * An opaque pagination token returned from the previous List operation in this request. If present, this indicates where to continue the listing.
     */
    NextToken?: NextToken;
    /**
     * List of SynthesisTask objects that provides information from the specified task in the list request, including output format, creation time, task status, and so on.
     */
    SynthesisTasks?: SynthesisTasks;
  }
  export type MaxResults = number;
  export type NextToken = string;
  export type OutputFormat = "json"|"mp3"|"ogg_vorbis"|"pcm"|string;
  export type OutputS3BucketName = string;
  export type OutputS3KeyPrefix = string;
  export type OutputUri = string;
  export interface PutLexiconInput {
    /**
     * Name of the lexicon. The name must follow the regular express format [0-9A-Za-z]{1,20}. That is, the name is a case-sensitive alphanumeric string up to 20 characters long. 
     */
    Name: LexiconName;
    /**
     * Content of the PLS lexicon as string data.
     */
    Content: LexiconContent;
  }
  export interface PutLexiconOutput {
  }
  export type RequestCharacters = number;
  export type SampleRate = string;
  export type Size = number;
  export type SnsTopicArn = string;
  export type SpeechMarkType = "sentence"|"ssml"|"viseme"|"word"|string;
  export type SpeechMarkTypeList = SpeechMarkType[];
  export interface StartSpeechSynthesisTaskInput {
    /**
     * Specifies the engine (standard or neural) for Amazon Polly to use when processing input text for speech synthesis. Using a voice that is not supported for the engine selected will result in an error.
     */
    Engine?: Engine;
    /**
     * Optional language code for the Speech Synthesis request. This is only necessary if using a bilingual voice, such as Aditi, which can be used for either Indian English (en-IN) or Hindi (hi-IN).  If a bilingual voice is used and no language code is specified, Amazon Polly will use the default language of the bilingual voice. The default language for any voice is the one returned by the DescribeVoices operation for the LanguageCode parameter. For example, if no language code is specified, Aditi will use Indian English rather than Hindi.
     */
    LanguageCode?: LanguageCode;
    /**
     * List of one or more pronunciation lexicon names you want the service to apply during synthesis. Lexicons are applied only if the language of the lexicon is the same as the language of the voice. 
     */
    LexiconNames?: LexiconNameList;
    /**
     * The format in which the returned output will be encoded. For audio stream, this will be mp3, ogg_vorbis, or pcm. For speech marks, this will be json. 
     */
    OutputFormat: OutputFormat;
    /**
     * Amazon S3 bucket name to which the output file will be saved.
     */
    OutputS3BucketName: OutputS3BucketName;
    /**
     * The Amazon S3 key prefix for the output speech file.
     */
    OutputS3KeyPrefix?: OutputS3KeyPrefix;
    /**
     * The audio frequency specified in Hz. The valid values for mp3 and ogg_vorbis are "8000", "16000", "22050", and "24000". The default value for standard voices is "22050". The default value for neural voices is "24000". Valid values for pcm are "8000" and "16000" The default value is "16000". 
     */
    SampleRate?: SampleRate;
    /**
     * ARN for the SNS topic optionally used for providing status notification for a speech synthesis task.
     */
    SnsTopicArn?: SnsTopicArn;
    /**
     * The type of speech marks returned for the input text.
     */
    SpeechMarkTypes?: SpeechMarkTypeList;
    /**
     * The input text to synthesize. If you specify ssml as the TextType, follow the SSML format for the input text. 
     */
    Text: Text;
    /**
     * Specifies whether the input text is plain text or SSML. The default value is plain text. 
     */
    TextType?: TextType;
    /**
     * Voice ID to use for the synthesis. 
     */
    VoiceId: VoiceId;
  }
  export interface StartSpeechSynthesisTaskOutput {
    /**
     * SynthesisTask object that provides information and attributes about a newly submitted speech synthesis task.
     */
    SynthesisTask?: SynthesisTask;
  }
  export interface SynthesisTask {
    /**
     * Specifies the engine (standard or neural) for Amazon Polly to use when processing input text for speech synthesis. Using a voice that is not supported for the engine selected will result in an error.
     */
    Engine?: Engine;
    /**
     * The Amazon Polly generated identifier for a speech synthesis task.
     */
    TaskId?: TaskId;
    /**
     * Current status of the individual speech synthesis task.
     */
    TaskStatus?: TaskStatus;
    /**
     * Reason for the current status of a specific speech synthesis task, including errors if the task has failed.
     */
    TaskStatusReason?: TaskStatusReason;
    /**
     * Pathway for the output speech file.
     */
    OutputUri?: OutputUri;
    /**
     * Timestamp for the time the synthesis task was started.
     */
    CreationTime?: DateTime;
    /**
     * Number of billable characters synthesized.
     */
    RequestCharacters?: RequestCharacters;
    /**
     * ARN for the SNS topic optionally used for providing status notification for a speech synthesis task.
     */
    SnsTopicArn?: SnsTopicArn;
    /**
     * List of one or more pronunciation lexicon names you want the service to apply during synthesis. Lexicons are applied only if the language of the lexicon is the same as the language of the voice. 
     */
    LexiconNames?: LexiconNameList;
    /**
     * The format in which the returned output will be encoded. For audio stream, this will be mp3, ogg_vorbis, or pcm. For speech marks, this will be json. 
     */
    OutputFormat?: OutputFormat;
    /**
     * The audio frequency specified in Hz. The valid values for mp3 and ogg_vorbis are "8000", "16000", "22050", and "24000". The default value for standard voices is "22050". The default value for neural voices is "24000". Valid values for pcm are "8000" and "16000" The default value is "16000". 
     */
    SampleRate?: SampleRate;
    /**
     * The type of speech marks returned for the input text.
     */
    SpeechMarkTypes?: SpeechMarkTypeList;
    /**
     * Specifies whether the input text is plain text or SSML. The default value is plain text. 
     */
    TextType?: TextType;
    /**
     * Voice ID to use for the synthesis. 
     */
    VoiceId?: VoiceId;
    /**
     * Optional language code for a synthesis task. This is only necessary if using a bilingual voice, such as Aditi, which can be used for either Indian English (en-IN) or Hindi (hi-IN).  If a bilingual voice is used and no language code is specified, Amazon Polly will use the default language of the bilingual voice. The default language for any voice is the one returned by the DescribeVoices operation for the LanguageCode parameter. For example, if no language code is specified, Aditi will use Indian English rather than Hindi.
     */
    LanguageCode?: LanguageCode;
  }
  export type SynthesisTasks = SynthesisTask[];
  export interface SynthesizeSpeechInput {
    /**
     * Specifies the engine (standard or neural) for Amazon Polly to use when processing input text for speech synthesis. Using a voice that is not supported for the engine selected will result in an error.
     */
    Engine?: Engine;
    /**
     * Optional language code for the Synthesize Speech request. This is only necessary if using a bilingual voice, such as Aditi, which can be used for either Indian English (en-IN) or Hindi (hi-IN).  If a bilingual voice is used and no language code is specified, Amazon Polly will use the default language of the bilingual voice. The default language for any voice is the one returned by the DescribeVoices operation for the LanguageCode parameter. For example, if no language code is specified, Aditi will use Indian English rather than Hindi.
     */
    LanguageCode?: LanguageCode;
    /**
     * List of one or more pronunciation lexicon names you want the service to apply during synthesis. Lexicons are applied only if the language of the lexicon is the same as the language of the voice. For information about storing lexicons, see PutLexicon.
     */
    LexiconNames?: LexiconNameList;
    /**
     *  The format in which the returned output will be encoded. For audio stream, this will be mp3, ogg_vorbis, or pcm. For speech marks, this will be json.  When pcm is used, the content returned is audio/pcm in a signed 16-bit, 1 channel (mono), little-endian format. 
     */
    OutputFormat: OutputFormat;
    /**
     * The audio frequency specified in Hz. The valid values for mp3 and ogg_vorbis are "8000", "16000", "22050", and "24000". The default value for standard voices is "22050". The default value for neural voices is "24000". Valid values for pcm are "8000" and "16000" The default value is "16000". 
     */
    SampleRate?: SampleRate;
    /**
     * The type of speech marks returned for the input text.
     */
    SpeechMarkTypes?: SpeechMarkTypeList;
    /**
     *  Input text to synthesize. If you specify ssml as the TextType, follow the SSML format for the input text. 
     */
    Text: Text;
    /**
     *  Specifies whether the input text is plain text or SSML. The default value is plain text. For more information, see Using SSML.
     */
    TextType?: TextType;
    /**
     *  Voice ID to use for the synthesis. You can get a list of available voice IDs by calling the DescribeVoices operation. 
     */
    VoiceId: VoiceId;
  }
  export interface SynthesizeSpeechOutput {
    /**
     *  Stream containing the synthesized speech. 
     */
    AudioStream?: AudioStream;
    /**
     *  Specifies the type audio stream. This should reflect the OutputFormat parameter in your request.     If you request mp3 as the OutputFormat, the ContentType returned is audio/mpeg.     If you request ogg_vorbis as the OutputFormat, the ContentType returned is audio/ogg.     If you request pcm as the OutputFormat, the ContentType returned is audio/pcm in a signed 16-bit, 1 channel (mono), little-endian format.    If you request json as the OutputFormat, the ContentType returned is audio/json.    
     */
    ContentType?: ContentType;
    /**
     * Number of characters synthesized.
     */
    RequestCharacters?: RequestCharacters;
  }
  export type TaskId = string;
  export type TaskStatus = "scheduled"|"inProgress"|"completed"|"failed"|string;
  export type TaskStatusReason = string;
  export type Text = string;
  export type TextType = "ssml"|"text"|string;
  export interface Voice {
    /**
     * Gender of the voice.
     */
    Gender?: Gender;
    /**
     * Amazon Polly assigned voice ID. This is the ID that you specify when calling the SynthesizeSpeech operation.
     */
    Id?: VoiceId;
    /**
     * Language code of the voice.
     */
    LanguageCode?: LanguageCode;
    /**
     * Human readable name of the language in English.
     */
    LanguageName?: LanguageName;
    /**
     * Name of the voice (for example, Salli, Kendra, etc.). This provides a human readable voice name that you might display in your application.
     */
    Name?: VoiceName;
    /**
     * Additional codes for languages available for the specified voice in addition to its default language.  For example, the default language for Aditi is Indian English (en-IN) because it was first used for that language. Since Aditi is bilingual and fluent in both Indian English and Hindi, this parameter would show the code hi-IN.
     */
    AdditionalLanguageCodes?: LanguageCodeList;
    /**
     * Specifies which engines (standard or neural) that are supported by a given voice.
     */
    SupportedEngines?: EngineList;
  }
  export type VoiceId = "Aditi"|"Amy"|"Astrid"|"Bianca"|"Brian"|"Camila"|"Carla"|"Carmen"|"Celine"|"Chantal"|"Conchita"|"Cristiano"|"Dora"|"Emma"|"Enrique"|"Ewa"|"Filiz"|"Geraint"|"Giorgio"|"Gwyneth"|"Hans"|"Ines"|"Ivy"|"Jacek"|"Jan"|"Joanna"|"Joey"|"Justin"|"Karl"|"Kendra"|"Kimberly"|"Lea"|"Liv"|"Lotte"|"Lucia"|"Lupe"|"Mads"|"Maja"|"Marlene"|"Mathieu"|"Matthew"|"Maxim"|"Mia"|"Miguel"|"Mizuki"|"Naja"|"Nicole"|"Penelope"|"Raveena"|"Ricardo"|"Ruben"|"Russell"|"Salli"|"Seoyeon"|"Takumi"|"Tatyana"|"Vicki"|"Vitoria"|"Zeina"|"Zhiyu"|string;
  export type VoiceList = Voice[];
  export type VoiceName = string;
  /**
   * A string in YYYY-MM-DD format that represents the latest possible API version that can be used in this service. Specify 'latest' to use the latest possible version.
   */
  export type apiVersion = "2016-06-10"|"latest"|string;
  export interface ClientApiVersions {
    /**
     * A string in YYYY-MM-DD format that represents the latest possible API version that can be used in this service. Specify 'latest' to use the latest possible version.
     */
    apiVersion?: apiVersion;
  }
  export type ClientConfiguration = ServiceConfigurationOptions & ClientApiVersions;
  /**
   * Contains interfaces for use with the Polly client.
   */
  export import Types = Polly;
}
export = Polly;