API - Corsair

Every elevenlabs.api.* operation is listed below with parameter shapes and return types from the plugin Zod schemas.

New to Corsair? See API access, authentication, and error handling.

Account

subscription

account.subscription The plan, credits used and limits. Risk: read

await corsair.elevenlabs.api.account.subscription({});

Input: empty object Output

Name	Type	Required	Description
`tier`	`string \| null`	No	—
`status`	`string \| null`	No	—
`character_count`	`number \| null`	No	—
`character_limit`	`number \| null`	No	—
`voice_slots_used`	`number \| null`	No	—
`voice_limit`	`number \| null`	No	—
`can_use_instant_voice_cloning`	`boolean \| null`	No	—
`can_use_professional_voice_cloning`	`boolean \| null`	No	—
`currency`	`string \| null`	No	—
`next_character_count_reset_unix`	`number \| null`	No	—

usageStats

account.usageStats Credits used over a time range, by product or model. Risk: read

await corsair.elevenlabs.api.account.usageStats({ startTime: 1793750400000, endTime: 1794355200000, groupBy: ['product_type'] });

Input

Name	Type	Required	Description
`startTime`	`number`	Yes	Start of the range, Unix time in milliseconds.
`endTime`	`number`	Yes	End of the range, Unix time in milliseconds.
`intervalSeconds`	`number`	No	Bucket size: 3600 hourly, 86400 daily. Default 86400.
`groupBy`	`string[]`	No	Columns to split by, for example product_type or model.
`timeZone`	`string`	No	IANA time zone for the buckets, for example Asia/Kolkata.

For workspaces billed in credits. One billed by cost gets a 400 naming a different report. Output

Name	Type	Required	Description
`columns`	`string[] \| null`	No	—
`column_types`	`string[] \| null`	No	—
`column_units`	`(string \| null)[] \| null`	No	—
`rows`	`any[][] \| null`	No	—

user

account.user The account’s user and workspace ids and its subscription. Key fields are removed from the result. Risk: read

await corsair.elevenlabs.api.account.user({});

Input: empty object The API reports the key (or part of it) here, and those fields are removed from the result. Output

Name	Type	Required	Description
`user_id`	`string \| null`	No	—
`workspace_id`	`string \| null`	No	—
`subscription`	`object \| null`	No	—
`first_name`	`string \| null`	No	—
`created_at`	`number \| null`	No	—
`seat_type`	`string \| null`	No	—

subscription full type

{
  tier?: string | null,
  status?: string | null,
  character_count?: number | null,
  character_limit?: number | null,
  voice_slots_used?: number | null,
  voice_limit?: number | null,
  can_use_instant_voice_cloning?: boolean | null,
  can_use_professional_voice_cloning?: boolean | null,
  currency?: string | null,
  next_character_count_reset_unix?: number | null
} | null

Agents

conversationDelete

agents.conversationDelete Delete a conversation, for good. Risk: destructive · Irreversible

await corsair.elevenlabs.api.agents.conversationDelete({ conversationId: '<conversation id>' });

Input

Name	Type	Required	Description
`conversationId`	`string`	Yes	conversation_id from agents.conversationsList.

Output: empty object

conversationGet

agents.conversationGet Read a conversation’s transcript, metadata and analysis. Risk: read

await corsair.elevenlabs.api.agents.conversationGet({ conversationId: '<conversation id>' });

Input

Name	Type	Required	Description
`conversationId`	`string`	Yes	conversation_id from agents.conversationsList.

Output

Name	Type	Required	Description
`conversation_id`	`string`	Yes	—
`agent_id`	`string \| null`	No	—
`status`	`string \| null`	No	—
`transcript`	`Record<string, any>[] \| null`	No	—
`metadata`	`Record<string, any> \| null`	No	—
`analysis`	`Record<string, any> \| null`	No	—
`has_audio`	`boolean \| null`	No	—

conversationsList

agents.conversationsList List conversations, optionally for one agent. Risk: read

await corsair.elevenlabs.api.agents.conversationsList({ agentId: '<agent id>', pageSize: 20 });

Input

Name	Type	Required	Description
`cursor`	`string`	No	—
`pageSize`	`number`	No	1 to 100. Default 30.
`agentId`	`string`	No	—
`search`	`string`	No	—
`summaryMode`	`exclude \| include`	No	—
`sortDirection`	`asc \| desc`	No	—

Output

Name	Type	Required	Description
`conversations`	`object[]`	Yes	—
`next_cursor`	`string \| null`	No	—
`has_more`	`boolean \| null`	No	—

conversations full type

{
  conversation_id: string,
  agent_id?: string | null,
  agent_name?: string | null,
  status?: string | null,
  start_time_unix_secs?: number | null,
  call_duration_secs?: number | null,
  message_count?: number | null
}[]

create

agents.create Create a conversational agent. Risk: write

await corsair.elevenlabs.api.agents.create({ name: 'Support', conversationConfig: { agent: { first_message: 'Hi, how can I help?', prompt: { prompt: 'You answer questions about orders.' } } } });

Input

Name	Type	Required	Description
`conversationConfig`	`Record<string, any>`	Yes	The agent’s configuration with the API’s field names, for example { agent: { first_message: “Hi”, prompt: { prompt: “You are…” } }, tts: { voice_id: ”…” } }.
`platformSettings`	`Record<string, any>`	No	—
`workflow`	`Record<string, any>`	No	—
`name`	`string`	No	—
`tags`	`string[]`	No	—

conversationConfig keeps the API’s field names and is passed through as sent. An empty agent is fine; agent.prompt.prompt and agent.first_message are the usual place to start. Output

Name	Type	Required	Description
`agent_id`	`string`	Yes	—

delete

agents.delete Delete an agent, for good. Risk: destructive · Irreversible

await corsair.elevenlabs.api.agents.delete({ agentId: '<agent id>' });

Input

Name	Type	Required	Description
`agentId`	`string`	Yes	Agent id, for example agent_3701k3ttaq12ewp8b7qv5rfyszkz.
`force`	`boolean`	No	Delete even if other agents use this one as a subagent.

Output: empty object

get

agents.get Read an agent’s full configuration. Risk: read

await corsair.elevenlabs.api.agents.get({ agentId: '<agent id>' });

Input

Name	Type	Required	Description
`agentId`	`string`	Yes	Agent id, for example agent_3701k3ttaq12ewp8b7qv5rfyszkz.
`versionId`	`string`	No	—
`branchId`	`string`	No	—

Output

Name	Type	Required	Description
`agent_id`	`string`	Yes	—
`name`	`string \| null`	No	—
`conversation_config`	`Record<string, any> \| null`	No	—
`platform_settings`	`Record<string, any> \| null`	No	—
`workflow`	`Record<string, any> \| null`	No	—
`metadata`	`Record<string, any> \| null`	No	—
`tags`	`string[] \| null`	No	—
`version_id`	`string \| null`	No	—
`branch_id`	`string \| null`	No	—

list

agents.list List conversational agents. Risk: read

await corsair.elevenlabs.api.agents.list({ pageSize: 20, search: 'support' });

Input

Name	Type	Required	Description
`cursor`	`string`	No	—
`pageSize`	`number`	No	1 to 100. Default 30.
`search`	`string`	No	—
`archived`	`boolean`	No	—
`showOnlyOwnedAgents`	`boolean`	No	—
`tags`	`string[]`	No	—
`sortBy`	`name \| created_at \| updated_at \| call_count_7d`	No	—
`sortDirection`	`asc \| desc`	No	—

Output

Name	Type	Required	Description
`agents`	`object[]`	Yes	—
`next_cursor`	`string \| null`	No	—
`has_more`	`boolean \| null`	No	—

agents full type

{
  agent_id: string,
  name?: string | null,
  tags?: string[] | null,
  created_at_unix_secs?: number | null,
  updated_at_unix_secs?: number | null,
  archived?: boolean | null
}[]

signedUrl

agents.signedUrl Mint a short-lived WebSocket URL a client can use to talk to an agent; the conversation spends credits. Risk: write

await corsair.elevenlabs.api.agents.signedUrl({ agentId: '<agent id>' });

Input

Name	Type	Required	Description
`agentId`	`string`	Yes	Agent id, for example agent_3701k3ttaq12ewp8b7qv5rfyszkz.
`includeConversationId`	`boolean`	No	—
`branchId`	`string`	No	—

The URL lets a client talk to the agent, and that conversation spends credits. Treat it as a secret. This plugin doesn’t hold the conversation itself. Output

Name	Type	Required	Description
`signed_url`	`string`	Yes	WebSocket URL a client can open to talk to the agent. Valid about 15 minutes; treat it as a secret.

update

agents.update Change an agent’s configuration, name or tags. Risk: write

await corsair.elevenlabs.api.agents.update({ agentId: '<agent id>', name: 'Support v2', tags: ['support'] });

Input

Name	Type	Required	Description
`agentId`	`string`	Yes	Agent id, for example agent_3701k3ttaq12ewp8b7qv5rfyszkz.
`conversationConfig`	`Record<string, any>`	No	—
`platformSettings`	`Record<string, any>`	No	—
`workflow`	`Record<string, any>`	No	—
`name`	`string`	No	—
`tags`	`string[]`	No	—
`versionDescription`	`string`	No	—
`branchId`	`string`	No	—

Output

Name	Type	Required	Description
`agent_id`	`string`	Yes	—
`name`	`string \| null`	No	—
`conversation_config`	`Record<string, any> \| null`	No	—
`platform_settings`	`Record<string, any> \| null`	No	—
`workflow`	`Record<string, any> \| null`	No	—
`metadata`	`Record<string, any> \| null`	No	—
`tags`	`string[] \| null`	No	—
`version_id`	`string \| null`	No	—
`branch_id`	`string \| null`	No	—

Audio Isolation

isolate

audioIsolation.isolate Remove background noise from audio and keep the voices; spends credits by audio length. The cleaned audio comes back in the result. Risk: write

await corsair.elevenlabs.api.audioIsolation.isolate({ fileBase64: '<base64 audio>', fileName: 'noisy.mp3', contentType: 'audio/mpeg' });

Input

Name	Type	Required	Description
`fileBase64`	`string`	Yes	The file, base64-encoded (at most 100 MB once decoded).
`fileName`	`string`	No	—
`contentType`	`string`	No	Media type of the file, for example audio/mpeg.
`fileFormat`	`pcm_s16le_16 \| other`	No	pcm_s16le_16 for raw 16-bit 16 kHz mono PCM, which is faster. Default other.

ElevenLabs refuses audio under about 4.6 s. No history item, so the output always comes back in the result. It takes no outputFormat, and a clip of about 7.5 s came back as 302,018 bytes in one run, so in a script hand on anything over about 18 s instead of returning it. Output

Name	Type	Required	Description
`contentBase64`	`string`	No	The output, base64-encoded. Missing when contentOmitted is set.
`contentType`	`string`	Yes	Media type of the output, for example audio/mpeg.
`byteLength`	`number`	Yes	Size of the output in bytes.
`contentOmitted`	`not_requested \| over_limit`	No	Set when contentBase64 was left out: includeContent was false, or the output was over the limit.
`note`	`string`	No	How to get output that was left out.
`songId`	`string`	No	The `song-id` header, on generated music.
`requestId`	`string`	No	The `request-id` header. Pass it as previousRequestIds for continuity between clips.
`traceId`	`string`	No	The `x-trace-id` header; quote it to ElevenLabs support.
`characterCost`	`number`	No	Credits charged, from the `character-cost` header. Music and stem separation don’t send it; account.usageStats has their cost.
`historyItemId`	`string`	No	The history item holding this audio. Read it again with history.getAudio. Only speech operations have one.

Dubbing

create

dubbing.create Start dubbing a video or audio file into another language; spends credits by length. Returns at once with a dubbing_id; follow it with dubbing.get until status is dubbed or failed. Risk: write

await corsair.elevenlabs.api.dubbing.create({ sourceUrl: 'https://example.com/clip.mp4', targetLang: 'es', watermark: true });

Input

Name	Type	Required	Description
`fileBase64`	`string`	No	The file, base64-encoded (at most 100 MB once decoded).
`fileName`	`string`	No	—
`contentType`	`string`	No	Media type of the file, for example audio/mpeg.
`sourceUrl`	`string`	No	URL of the video or audio, instead of fileBase64.
`targetLang`	`string`	Yes	Language to dub into, as an ISO 639 code.
`sourceLang`	`string`	No	Default auto.
`name`	`string`	No	—
`numSpeakers`	`number`	No	0 (default) detects them.
`watermark`	`boolean`	No	—
`startTime`	`number`	No	—
`endTime`	`number`	No	—
`highestResolution`	`boolean`	No	—
`dropBackgroundAudio`	`boolean`	No	—
`useProfanityFilter`	`boolean`	No	—
`disableVoiceCloning`	`boolean`	No	true uses a similar voice from the library instead of cloning the speakers in the source.

Returns at once. Follow dubbing_id with dubbing.get until status is dubbed or failed. By default dubbing clones the speakers it hears; set disableVoiceCloning true to use library voices instead. Output

Name	Type	Required	Description
`dubbing_id`	`string`	Yes	—
`expected_duration_sec`	`number \| null`	No	—

delete

dubbing.delete Delete a dub, for good. Risk: destructive · Irreversible

await corsair.elevenlabs.api.dubbing.delete({ dubbingId: '<dubbing id>' });

Input

Name	Type	Required	Description
`dubbingId`	`string`	Yes	The dubbing_id that dubbing.create returned.

Output

Name	Type	Required	Description
`status`	`string \| null`	No	—

get

dubbing.get Where a dub stands: status dubbed or failed are final. Risk: read

await corsair.elevenlabs.api.dubbing.get({ dubbingId: '<dubbing id>' });

Input

Name	Type	Required	Description
`dubbingId`	`string`	Yes	The dubbing_id that dubbing.create returned.

Output

Name	Type	Required	Description
`dubbing_id`	`string`	Yes	—
`name`	`string \| null`	No	—
`status`	`string`	Yes	Where the dub stands: queued, preparing, dubbing, dubbed, failed.
`source_language`	`string \| null`	No	—
`target_languages`	`string[] \| null`	No	—
`editable`	`boolean \| null`	No	—
`created_at`	`string \| null`	No	—
`media_metadata`	`Record<string, any> \| null`	No	—
`error`	`string \| null`	No	—

getAudio

dubbing.getAudio Download a finished dub as base64 (audio or video), subject to the same size limit as other audio. Risk: read

await corsair.elevenlabs.api.dubbing.getAudio({ dubbingId: '<dubbing id>', languageCode: 'es', maxBytes: 10000000 });

Input

Name	Type	Required	Description
`dubbingId`	`string`	Yes	The dubbing_id that dubbing.create returned.
`languageCode`	`string`	Yes	A target language of the dub; for transcripts also `source`.
`includeContent`	`boolean`	No	false returns the size and type only. Default true.
`maxBytes`	`number`	No	Largest audio to carry in the result, in bytes. Over it, contentBase64 is left out and contentOmitted says so. Default 700000 (about 43 s of 128 kbps mp3), so the base64 fits the 1,000,000 byte cap on a script’s result on Corsair’s cloud. At most 50 MB.

Audio or video, per contentType. Subject to maxBytes, so raise it for anything long. Output

Name	Type	Required	Description
`contentBase64`	`string`	No	The output, base64-encoded. Missing when contentOmitted is set.
`contentType`	`string`	Yes	Media type of the output, for example audio/mpeg.
`byteLength`	`number`	Yes	Size of the output in bytes.
`contentOmitted`	`not_requested \| over_limit`	No	Set when contentBase64 was left out: includeContent was false, or the output was over the limit.
`note`	`string`	No	How to get output that was left out.
`songId`	`string`	No	The `song-id` header, on generated music.
`requestId`	`string`	No	The `request-id` header. Pass it as previousRequestIds for continuity between clips.
`traceId`	`string`	No	The `x-trace-id` header; quote it to ElevenLabs support.
`characterCost`	`number`	No	Credits charged, from the `character-cost` header. Music and stem separation don’t send it; account.usageStats has their cost.
`historyItemId`	`string`	No	The history item holding this audio. Read it again with history.getAudio. Only speech operations have one.

getTranscript

dubbing.getTranscript Read a dub’s transcript as srt, webvtt or json. Risk: read

await corsair.elevenlabs.api.dubbing.getTranscript({ dubbingId: '<dubbing id>', languageCode: 'es', formatType: 'srt' });

Input

Name	Type	Required	Description
`dubbingId`	`string`	Yes	The dubbing_id that dubbing.create returned.
`languageCode`	`string`	Yes	A target language of the dub; for transcripts also `source`.
`formatType`	`srt \| webvtt \| json`	No	Default srt.

Output

Name	Type	Required	Description
`transcript_format`	`string \| null`	No	—
`srt`	`string \| null`	No	—
`webvtt`	`string \| null`	No	—
`json`	`Record<string, any> \| null`	No	—

list

dubbing.list List dubs, newest first by default. Risk: read

await corsair.elevenlabs.api.dubbing.list({ pageSize: 20, dubbingStatus: 'dubbed' });

Input

Name	Type	Required	Description
`cursor`	`string`	No	—
`pageSize`	`number`	No	1 to 200. Default 100.
`dubbingStatus`	`dubbing \| dubbed \| failed`	No	—
`filterByCreator`	`personal \| others \| all`	No	—
`orderBy`	`created_at \| name`	No	—
`orderDirection`	`DESCENDING \| ASCENDING`	No	—

Output

Name	Type	Required	Description
`dubs`	`object[]`	Yes	—
`next_cursor`	`string \| null`	No	—
`has_more`	`boolean \| null`	No	—

dubs full type

{
  dubbing_id: string,
  name?: string | null,
  status: string,
  source_language?: string | null,
  target_languages?: string[] | null,
  editable?: boolean | null,
  created_at?: string | null,
  media_metadata?: Record<string, any> | null,
  error?: string | null
}[]

Forced Alignment

create

forcedAlignment.create Line up known text with audio to get word and character timings; spends credits by audio length. Risk: write

await corsair.elevenlabs.api.forcedAlignment.create({ fileBase64: '<base64 audio>', fileName: 'line.mp3', contentType: 'audio/mpeg', text: 'Hello from Corsair.' });

Input

Name	Type	Required	Description
`fileBase64`	`string`	Yes	The file, base64-encoded (at most 100 MB once decoded).
`fileName`	`string`	No	—
`contentType`	`string`	No	Media type of the file, for example audio/mpeg.
`text`	`string`	Yes	The text spoken in the audio.

Output

Name	Type	Required	Description
`characters`	`object[] \| null`	No	—
`words`	`object[] \| null`	No	—
`loss`	`number \| null`	No	—

characters full type

{
  text?: string | null,
  start?: number | null,
  end?: number | null
}[] | null

words full type

{
  text?: string | null,
  start?: number | null,
  end?: number | null
}[] | null

History

delete

history.delete Delete a history item and its audio, for good. Risk: destructive · Irreversible

await corsair.elevenlabs.api.history.delete({ historyItemId: '<history item id>' });

Input

Name	Type	Required	Description
`historyItemId`	`string`	Yes	historyItemId from a speech operation, or history_item_id from history.list.

Output

Name	Type	Required	Description
`status`	`string \| null`	No	—

get

history.get Read one history item: its text, voice, model and cost. Risk: read

await corsair.elevenlabs.api.history.get({ historyItemId: '<history item id>' });

Input

Name	Type	Required	Description
`historyItemId`	`string`	Yes	historyItemId from a speech operation, or history_item_id from history.list.

Output

Name	Type	Required	Description
`history_item_id`	`string`	Yes	—
`request_id`	`string \| null`	No	—
`voice_id`	`string \| null`	No	—
`model_id`	`string \| null`	No	—
`voice_name`	`string \| null`	No	—
`voice_category`	`string \| null`	No	—
`text`	`string \| null`	No	—
`date_unix`	`number \| null`	No	—
`character_count_change_from`	`number \| null`	No	—
`character_count_change_to`	`number \| null`	No	—
`content_type`	`string \| null`	No	—
`state`	`string \| null`	No	—
`source`	`string \| null`	No	—
`output_format`	`string \| null`	No	—

getAudio

history.getAudio Download a history item’s audio as base64, subject to a size limit you can raise with maxBytes. Risk: read

await corsair.elevenlabs.api.history.getAudio({ historyItemId: '<history item id>', maxBytes: 5000000 });

Input

Name	Type	Required	Description
`historyItemId`	`string`	Yes	historyItemId from a speech operation, or history_item_id from history.list.
`includeContent`	`boolean`	No	false returns the size and type only. Default true.
`maxBytes`	`number`	No	Largest audio to carry in the result, in bytes. Over it, contentBase64 is left out and contentOmitted says so. Default 700000 (about 43 s of 128 kbps mp3), so the base64 fits the 1,000,000 byte cap on a script’s result on Corsair’s cloud. At most 50 MB.

Subject to maxBytes (default 700000 bytes), so raise it for anything long. A new item can take a few seconds to appear, and reading it before then answers 400. Output

Name	Type	Required	Description
`contentBase64`	`string`	No	The output, base64-encoded. Missing when contentOmitted is set.
`contentType`	`string`	Yes	Media type of the output, for example audio/mpeg.
`byteLength`	`number`	Yes	Size of the output in bytes.
`contentOmitted`	`not_requested \| over_limit`	No	Set when contentBase64 was left out: includeContent was false, or the output was over the limit.
`note`	`string`	No	How to get output that was left out.
`songId`	`string`	No	The `song-id` header, on generated music.
`requestId`	`string`	No	The `request-id` header. Pass it as previousRequestIds for continuity between clips.
`traceId`	`string`	No	The `x-trace-id` header; quote it to ElevenLabs support.
`characterCost`	`number`	No	Credits charged, from the `character-cost` header. Music and stem separation don’t send it; account.usageStats has their cost.
`historyItemId`	`string`	No	The history item holding this audio. Read it again with history.getAudio. Only speech operations have one.

list

history.list List generated speech, newest first by default. Risk: read

await corsair.elevenlabs.api.history.list({ pageSize: 20, source: 'TTS' });

Input

Name	Type	Required	Description
`pageSize`	`number`	No	1 to 1000. Default 100. Deleted items count, so a page can be short or empty while has_more is true.
`startAfterHistoryItemId`	`string`	No	last_history_item_id from the previous page.
`voiceId`	`string`	No	—
`modelId`	`string`	No	—
`dateBeforeUnix`	`number`	No	—
`dateAfterUnix`	`number`	No	—
`sortDirection`	`asc \| desc`	No	—
`search`	`string`	No	—
`source`	`TTS \| STS \| Flows`	No	—

Deleted items still count toward a page, so a page can be short or even empty while has_more is true. Keep going with last_history_item_id until has_more is false. Output

Name	Type	Required	Description
`history`	`object[]`	Yes	—
`last_history_item_id`	`string \| null`	No	—
`has_more`	`boolean \| null`	No	—
`scanned_until`	`number \| null`	No	—

history full type

{
  history_item_id: string,
  request_id?: string | null,
  voice_id?: string | null,
  model_id?: string | null,
  voice_name?: string | null,
  voice_category?: string | null,
  text?: string | null,
  date_unix?: number | null,
  character_count_change_from?: number | null,
  character_count_change_to?: number | null,
  content_type?: string | null,
  state?: string | null,
  source?: string | null,
  output_format?: string | null
}[]

Models

list

models.list List available models with their languages and rates. Risk: read

await corsair.elevenlabs.api.models.list({});

Input: empty object Output

Name	Type	Required	Description
`models`	`object[]`	Yes	—

models full type

{
  model_id: string,
  name?: string | null,
  description?: string | null,
  can_do_text_to_speech?: boolean | null,
  can_do_voice_conversion?: boolean | null,
  maximum_text_length_per_request?: number | null,
  languages?: Record<string, any>[] | null,
  model_rates?: Record<string, any> | null
}[]

Music

compose

music.compose Generate a song from a prompt or a composition plan; spends credits by length. ElevenLabs keeps no copy, so the audio comes back in the result. At the default format, about 24 KB/s, a song over about 31 s is too big to return from a script on Corsair’s cloud, whose cap is 1,000,000 bytes: pick a lower-bitrate outputFormat, or pass contentBase64 to the next operation and return only metadata. Risk: write

await corsair.elevenlabs.api.music.compose({ prompt: 'a calm piano loop', musicLengthMs: 10000, forceInstrumental: true, outputFormat: 'mp3_22050_32' });

Input

Name	Type	Required	Description
`prompt`	`string`	No	What the song should be.
`compositionPlan`	`Record<string, any>`	No	A plan from music.plan, instead of prompt.
`musicLengthMs`	`number`	No	3 s to 10 min. Only with prompt.
`lyricsText`	`string`	No	—
`generationMode`	`track \| loop \| ambience \| video_to_music`	No	—
`modelId`	`string`	No	Model id. Default music_v2_5.
`seed`	`number`	No	—
`forceInstrumental`	`boolean`	No	—
`outputFormat`	`string`	No	codec_samplerate_bitrate, for example mp3_22050_32 or mp3_44100_128 (the usual default). Higher quality formats may need a higher plan.

Pass prompt or compositionPlan; musicLengthMs only works with prompt. Defaults to music_v2_5, because the API’s own default, music_v1, is removed on 2027-01-31. No characterCost header; account.usageStats has the cost. Output

Name	Type	Required	Description
`contentBase64`	`string`	No	The output, base64-encoded. Missing when contentOmitted is set.
`contentType`	`string`	Yes	Media type of the output, for example audio/mpeg.
`byteLength`	`number`	Yes	Size of the output in bytes.
`contentOmitted`	`not_requested \| over_limit`	No	Set when contentBase64 was left out: includeContent was false, or the output was over the limit.
`note`	`string`	No	How to get output that was left out.
`songId`	`string`	No	The `song-id` header, on generated music.
`requestId`	`string`	No	The `request-id` header. Pass it as previousRequestIds for continuity between clips.
`traceId`	`string`	No	The `x-trace-id` header; quote it to ElevenLabs support.
`characterCost`	`number`	No	Credits charged, from the `character-cost` header. Music and stem separation don’t send it; account.usageStats has their cost.
`historyItemId`	`string`	No	The history item holding this audio. Read it again with history.getAudio. Only speech operations have one.

plan

music.plan Draft a composition plan from a prompt, to edit and pass to music.compose. Risk: write

await corsair.elevenlabs.api.music.plan({ prompt: 'an upbeat jingle', musicLengthMs: 15000 });

Input

Name	Type	Required	Description
`prompt`	`string`	Yes	What the song should be.
`musicLengthMs`	`number`	No	—
`sourceCompositionPlan`	`Record<string, any>`	No	—
`modelId`	`string`	No	Model id. Default music_v2_5.

The API’s schema allows chunks or sections in a plan; music_v2_5 answered with chunks in our test. Pass the plan back to music.compose as compositionPlan. Output

Name	Type	Required	Description
`chunks`	`Record<string, any>[] \| null`	No	—
`sections`	`Record<string, any>[] \| null`	No	—
`positive_global_styles`	`string[] \| null`	No	—
`negative_global_styles`	`string[] \| null`	No	—

separateStems

music.separateStems Split a song into stems, returned as a zip; spends credits by audio length. Risk: write

await corsair.elevenlabs.api.music.separateStems({ fileBase64: '<base64 song>', fileName: 'song.mp3', stemVariationId: 'two_stems_v1' });

Input

Name	Type	Required	Description
`fileBase64`	`string`	Yes	The file, base64-encoded (at most 100 MB once decoded).
`fileName`	`string`	No	—
`contentType`	`string`	No	Media type of the file, for example audio/mpeg.
`stemVariationId`	`two_stems_v1 \| six_stems_v1`	No	Two stems (vocals, rest) or six. Default six_stems_v1.
`outputFormat`	`string`	No	codec_samplerate_bitrate, for example mp3_22050_32 or mp3_44100_128 (the usual default). Higher quality formats may need a higher plan.

Returns a zip with one file per stem. No characterCost header; account.usageStats has the cost. Output

Name	Type	Required	Description
`contentBase64`	`string`	No	The output, base64-encoded. Missing when contentOmitted is set.
`contentType`	`string`	Yes	Media type of the output, for example audio/mpeg.
`byteLength`	`number`	Yes	Size of the output in bytes.
`contentOmitted`	`not_requested \| over_limit`	No	Set when contentBase64 was left out: includeContent was false, or the output was over the limit.
`note`	`string`	No	How to get output that was left out.
`songId`	`string`	No	The `song-id` header, on generated music.
`requestId`	`string`	No	The `request-id` header. Pass it as previousRequestIds for continuity between clips.
`traceId`	`string`	No	The `x-trace-id` header; quote it to ElevenLabs support.
`characterCost`	`number`	No	Credits charged, from the `character-cost` header. Music and stem separation don’t send it; account.usageStats has their cost.
`historyItemId`	`string`	No	The history item holding this audio. Read it again with history.getAudio. Only speech operations have one.

Single Use Tokens

create

singleUseTokens.create Mint a single-use token, valid 15 minutes, to give a client instead of the API key. Risk: write

await corsair.elevenlabs.api.singleUseTokens.create({ tokenType: 'tts_websocket' });

Input

Name	Type	Required	Description
`tokenType`	`realtime_scribe \| batch_scribe \| tts_websocket`	Yes	—

Works once and expires after 15 minutes. Treat it as a secret until it’s used. Output

Name	Type	Required	Description
`token`	`string`	Yes	Expires after 15 minutes and works once. Hand it to a client instead of the API key.

Sound Effects

generate

soundEffects.generate Generate a sound effect from a description; spends credits by duration. ElevenLabs keeps no copy, so the audio comes back in the result. Risk: write

await corsair.elevenlabs.api.soundEffects.generate({ text: 'a door creaking open', durationSeconds: 2, outputFormat: 'mp3_22050_32' });

Input

Name	Type	Required	Description
`text`	`string`	Yes	What the sound should be.
`durationSeconds`	`number`	No	0.5 to 30 s. Left out, the model picks a length.
`loop`	`boolean`	No	—
`promptInfluence`	`number`	No	—
`modelId`	`string`	No	Model id. Default eleven_text_to_sound_v3.
`outputFormat`	`string`	No	codec_samplerate_bitrate, for example mp3_22050_32 or mp3_44100_128 (the usual default). Higher quality formats may need a higher plan.

Defaults to eleven_text_to_sound_v3, because the API’s own default, eleven_text_to_sound_v2, is removed on 2027-03-01. Output

Name	Type	Required	Description
`contentBase64`	`string`	No	The output, base64-encoded. Missing when contentOmitted is set.
`contentType`	`string`	Yes	Media type of the output, for example audio/mpeg.
`byteLength`	`number`	Yes	Size of the output in bytes.
`contentOmitted`	`not_requested \| over_limit`	No	Set when contentBase64 was left out: includeContent was false, or the output was over the limit.
`note`	`string`	No	How to get output that was left out.
`songId`	`string`	No	The `song-id` header, on generated music.
`requestId`	`string`	No	The `request-id` header. Pass it as previousRequestIds for continuity between clips.
`traceId`	`string`	No	The `x-trace-id` header; quote it to ElevenLabs support.
`characterCost`	`number`	No	Credits charged, from the `character-cost` header. Music and stem separation don’t send it; account.usageStats has their cost.
`historyItemId`	`string`	No	The history item holding this audio. Read it again with history.getAudio. Only speech operations have one.

Speech To Text

convert

speechToText.convert Transcribe audio or video from a URL or a base64 file; spends credits by audio length. With webhook true it returns at once and delivers the transcript by webhook. Risk: write

await corsair.elevenlabs.api.speechToText.convert({ sourceUrl: 'https://example.com/call.mp3', diarize: true });

Input

Name	Type	Required	Description
`fileBase64`	`string`	No	The file, base64-encoded (at most 100 MB once decoded).
`fileName`	`string`	No	—
`contentType`	`string`	No	Media type of the file, for example audio/mpeg.
`sourceUrl`	`string`	No	URL of the audio or video to transcribe, instead of fileBase64: a hosted file or a page ElevenLabs can read, such as YouTube.
`modelId`	`string`	No	Model id. Default scribe_v2.
`languageCode`	`string`	No	—
`tagAudioEvents`	`boolean`	No	—
`numSpeakers`	`number`	No	—
`timestampsGranularity`	`none \| word \| character`	No	—
`diarize`	`boolean`	No	—
`keyterms`	`string[]`	No	—
`temperature`	`number`	No	—
`seed`	`number`	No	—
`fileFormat`	`pcm_s16le_16 \| other`	No	pcm_s16le_16 for raw 16-bit 16 kHz mono PCM, which is faster. Default other.
`useMultiChannel`	`boolean`	No	—
`webhook`	`boolean`	No	true returns at once with request_id and transcription_id, and delivers the transcript to the workspace’s speech-to-text webhooks (the speechToText.transcriptionCompleted event). Needs a webhook set up in ElevenLabs.
`webhookId`	`string`	No	Send to this webhook only. Needs webhook: true.
`webhookMetadata`	`Record<string, any>`	No	Sent back as webhook_metadata in the webhook event. Needs webhook: true.
`enableLogging`	`boolean`	No	false turns on zero retention (enterprise plans): no history item, so the audio can only come back in this result.

Pass fileBase64 or sourceUrl. With webhook true it returns request_id and transcription_id at once and delivers to the workspace’s speech-to-text webhooks, which must exist first. speechToText.getTranscript answers 404 until the transcript is ready, the same as for a wrong id, which is why this isn’t declared as a job. Output

Name	Type	Required	Description
`language_code`	`string \| null`	No	—
`language_probability`	`number \| null`	No	—
`text`	`string \| null`	No	—
`words`	`Record<string, any>[] \| null`	No	—
`transcription_id`	`string \| null`	No	—
`audio_duration_secs`	`number \| null`	No	—
`transcripts`	`object[] \| null`	No	One transcript per channel, when useMultiChannel is on.
`message`	`string \| null`	No	Set in webhook mode.
`request_id`	`string \| null`	No	Set in webhook mode.

transcripts full type

{
  language_code?: string | null,
  language_probability?: number | null,
  text?: string | null,
  words?: Record<string, any>[] | null,
  transcription_id?: string | null,
  audio_duration_secs?: number | null
}[] | null

deleteTranscript

speechToText.deleteTranscript Delete a stored transcript, for good. Risk: destructive · Irreversible

await corsair.elevenlabs.api.speechToText.deleteTranscript({ transcriptionId: '<transcription id>' });

Input

Name	Type	Required	Description
`transcriptionId`	`string`	Yes	The transcription_id that speechToText.convert returned.

Output

Name	Type	Required	Description
`status`	`string \| null`	No	—

getTranscript

speechToText.getTranscript Read a transcript by the transcription_id speechToText.convert returned. Risk: read

await corsair.elevenlabs.api.speechToText.getTranscript({ transcriptionId: '<transcription id>' });

Input

Name	Type	Required	Description
`transcriptionId`	`string`	Yes	The transcription_id that speechToText.convert returned.

Output

Name	Type	Required	Description
`language_code`	`string \| null`	No	—
`language_probability`	`number \| null`	No	—
`text`	`string \| null`	No	—
`words`	`Record<string, any>[] \| null`	No	—
`transcription_id`	`string \| null`	No	—
`audio_duration_secs`	`number \| null`	No	—
`transcripts`	`object[] \| null`	No	—

transcripts full type

{
  language_code?: string | null,
  language_probability?: number | null,
  text?: string | null,
  words?: Record<string, any>[] | null,
  transcription_id?: string | null,
  audio_duration_secs?: number | null
}[] | null

Text To Dialogue

convert

textToDialogue.convert Turn lines spoken by several voices into one audio clip; spends credits by text length. Risk: write

await corsair.elevenlabs.api.textToDialogue.convert({ inputs: [{ text: 'Are you ready?', voiceId: 'FGY2WhTYpPnrIDTdsKH5' }, { text: 'Always.', voiceId: 'JBFqnCBsd6RMkjVDRZzb' }], outputFormat: 'mp3_22050_32' });

Input

Name	Type	Required	Description
`inputs`	`object[]`	Yes	The lines in order, each with the voice that speaks it.
`modelId`	`string`	No	Model id. Default eleven_v3.
`languageCode`	`string`	No	—
`settings`	`Record<string, any>`	No	Dialogue settings (stability and so on), with the API’s field names.
`seed`	`number`	No	—
`previousText`	`string`	No	—
`futureText`	`string`	No	—
`pronunciationDictionaryLocators`	`object[]`	No	—
`applyTextNormalization`	`auto \| on \| off`	No	—
`outputFormat`	`string`	No	codec_samplerate_bitrate, for example mp3_22050_32 or mp3_44100_128 (the usual default). Higher quality formats may need a higher plan.
`enableLogging`	`boolean`	No	false turns on zero retention (enterprise plans): no history item, so the audio can only come back in this result.
`includeContent`	`boolean`	No	false leaves contentBase64 out of the result and returns its metadata only, which keeps a long clip out of an agent’s context. Read the audio later with history.getAudio. Default true.
`maxBytes`	`number`	No	Largest audio to carry in the result, in bytes. Over it, contentBase64 is left out and contentOmitted says so. Default 700000 (about 43 s of 128 kbps mp3), so the base64 fits the 1,000,000 byte cap on a script’s result on Corsair’s cloud. At most 50 MB.

inputs full type

{
  text: string,
  voiceId: string
}[]

pronunciationDictionaryLocators full type

{
  pronunciation_dictionary_id: string,
  version_id?: string | null
}[]

Output

Name	Type	Required	Description
`contentBase64`	`string`	No	The output, base64-encoded. Missing when contentOmitted is set.
`contentType`	`string`	Yes	Media type of the output, for example audio/mpeg.
`byteLength`	`number`	Yes	Size of the output in bytes.
`contentOmitted`	`not_requested \| over_limit`	No	Set when contentBase64 was left out: includeContent was false, or the output was over the limit.
`note`	`string`	No	How to get output that was left out.
`songId`	`string`	No	The `song-id` header, on generated music.
`requestId`	`string`	No	The `request-id` header. Pass it as previousRequestIds for continuity between clips.
`traceId`	`string`	No	The `x-trace-id` header; quote it to ElevenLabs support.
`characterCost`	`number`	No	Credits charged, from the `character-cost` header. Music and stem separation don’t send it; account.usageStats has their cost.
`historyItemId`	`string`	No	The history item holding this audio. Read it again with history.getAudio. Only speech operations have one.

Text To Speech

convert

textToSpeech.convert Turn text into speech; spends credits by text length. Returns the audio as base64 with its history item id; set includeContent false to keep a long clip out of the result and read it later with history.getAudio. Risk: write

await corsair.elevenlabs.api.textToSpeech.convert({ voiceId: 'FGY2WhTYpPnrIDTdsKH5', text: 'Your order has shipped.', modelId: 'eleven_flash_v2_5', outputFormat: 'mp3_22050_32' });

Input

Name	Type	Required	Description
`voiceId`	`string`	Yes	Voice id, for example FGY2WhTYpPnrIDTdsKH5. Find ids with voices.list.
`text`	`string`	Yes	The text to speak.
`modelId`	`string`	No	Model id from models.list, for example eleven_flash_v2_5 (fast and cheap) or eleven_multilingual_v2 (the default).
`languageCode`	`string`	No	—
`voiceSettings`	`object`	No	Overrides the voice’s stored settings for this call. Fields keep the API’s names.
`pronunciationDictionaryLocators`	`object[]`	No	—
`seed`	`number`	No	—
`previousText`	`string`	No	—
`nextText`	`string`	No	—
`previousRequestIds`	`string[]`	No	—
`nextRequestIds`	`string[]`	No	—
`applyTextNormalization`	`auto \| on \| off`	No	—
`applyLanguageTextNormalization`	`boolean`	No	—
`outputFormat`	`string`	No	codec_samplerate_bitrate, for example mp3_22050_32 or mp3_44100_128 (the usual default). Higher quality formats may need a higher plan.
`enableLogging`	`boolean`	No	false turns on zero retention (enterprise plans): no history item, so the audio can only come back in this result.
`includeContent`	`boolean`	No	false leaves contentBase64 out of the result and returns its metadata only, which keeps a long clip out of an agent’s context. Read the audio later with history.getAudio. Default true.
`maxBytes`	`number`	No	Largest audio to carry in the result, in bytes. Over it, contentBase64 is left out and contentOmitted says so. Default 700000 (about 43 s of 128 kbps mp3), so the base64 fits the 1,000,000 byte cap on a script’s result on Corsair’s cloud. At most 50 MB.

Spends credits by text length (characterCost says how many). The history item keeps its own stored format, so history.getAudio can return a different bitrate from outputFormat.

voiceSettings full type

{
  stability?: number | null,
  similarity_boost?: number | null,
  style?: number | null,
  use_speaker_boost?: boolean | null,
  speed?: number | null
}

pronunciationDictionaryLocators full type

{
  pronunciation_dictionary_id: string,
  version_id?: string | null
}[]

Output

Name	Type	Required	Description
`contentBase64`	`string`	No	The output, base64-encoded. Missing when contentOmitted is set.
`contentType`	`string`	Yes	Media type of the output, for example audio/mpeg.
`byteLength`	`number`	Yes	Size of the output in bytes.
`contentOmitted`	`not_requested \| over_limit`	No	Set when contentBase64 was left out: includeContent was false, or the output was over the limit.
`note`	`string`	No	How to get output that was left out.
`songId`	`string`	No	The `song-id` header, on generated music.
`requestId`	`string`	No	The `request-id` header. Pass it as previousRequestIds for continuity between clips.
`traceId`	`string`	No	The `x-trace-id` header; quote it to ElevenLabs support.
`characterCost`	`number`	No	Credits charged, from the `character-cost` header. Music and stem separation don’t send it; account.usageStats has their cost.
`historyItemId`	`string`	No	The history item holding this audio. Read it again with history.getAudio. Only speech operations have one.

convertWithTimestamps

textToSpeech.convertWithTimestamps Turn text into speech with the start and end time of each character; spends credits by text length. Same audio fields as textToSpeech.convert. Risk: write

await corsair.elevenlabs.api.textToSpeech.convertWithTimestamps({ voiceId: 'FGY2WhTYpPnrIDTdsKH5', text: 'Hi there.', modelId: 'eleven_flash_v2_5' });

Input

Name	Type	Required	Description
`voiceId`	`string`	Yes	Voice id, for example FGY2WhTYpPnrIDTdsKH5. Find ids with voices.list.
`text`	`string`	Yes	The text to speak.
`modelId`	`string`	No	Model id from models.list, for example eleven_flash_v2_5 (fast and cheap) or eleven_multilingual_v2 (the default).
`languageCode`	`string`	No	—
`voiceSettings`	`object`	No	Overrides the voice’s stored settings for this call. Fields keep the API’s names.
`pronunciationDictionaryLocators`	`object[]`	No	—
`seed`	`number`	No	—
`previousText`	`string`	No	—
`nextText`	`string`	No	—
`previousRequestIds`	`string[]`	No	—
`nextRequestIds`	`string[]`	No	—
`applyTextNormalization`	`auto \| on \| off`	No	—
`applyLanguageTextNormalization`	`boolean`	No	—
`outputFormat`	`string`	No	codec_samplerate_bitrate, for example mp3_22050_32 or mp3_44100_128 (the usual default). Higher quality formats may need a higher plan.
`enableLogging`	`boolean`	No	false turns on zero retention (enterprise plans): no history item, so the audio can only come back in this result.
`includeContent`	`boolean`	No	false leaves contentBase64 out of the result and returns its metadata only, which keeps a long clip out of an agent’s context. Read the audio later with history.getAudio. Default true.
`maxBytes`	`number`	No	Largest audio to carry in the result, in bytes. Over it, contentBase64 is left out and contentOmitted says so. Default 700000 (about 43 s of 128 kbps mp3), so the base64 fits the 1,000,000 byte cap on a script’s result on Corsair’s cloud. At most 50 MB.

Same audio fields as textToSpeech.convert, plus alignment and normalized_alignment.

voiceSettings full type

{
  stability?: number | null,
  similarity_boost?: number | null,
  style?: number | null,
  use_speaker_boost?: boolean | null,
  speed?: number | null
}

pronunciationDictionaryLocators full type

{
  pronunciation_dictionary_id: string,
  version_id?: string | null
}[]

Output

Name	Type	Required	Description
`contentBase64`	`string`	No	The output, base64-encoded. Missing when contentOmitted is set.
`contentType`	`string`	Yes	Media type of the output, for example audio/mpeg.
`byteLength`	`number`	Yes	Size of the output in bytes.
`contentOmitted`	`not_requested \| over_limit`	No	Set when contentBase64 was left out: includeContent was false, or the output was over the limit.
`note`	`string`	No	How to get output that was left out.
`songId`	`string`	No	The `song-id` header, on generated music.
`requestId`	`string`	No	The `request-id` header. Pass it as previousRequestIds for continuity between clips.
`traceId`	`string`	No	The `x-trace-id` header; quote it to ElevenLabs support.
`characterCost`	`number`	No	Credits charged, from the `character-cost` header. Music and stem separation don’t send it; account.usageStats has their cost.
`historyItemId`	`string`	No	The history item holding this audio. Read it again with history.getAudio. Only speech operations have one.
`alignment`	`object \| null`	No	When each character of the text starts and ends, in seconds.
`normalized_alignment`	`object \| null`	No	The same for the text after normalization, which is what was spoken.

alignment full type

{
  characters?: string[] | null,
  character_start_times_seconds?: number[] | null,
  character_end_times_seconds?: number[] | null
} | null

normalized_alignment full type

{
  characters?: string[] | null,
  character_start_times_seconds?: number[] | null,
  character_end_times_seconds?: number[] | null
} | null

Voice Changer

convert

voiceChanger.convert Say the same thing in a different voice, from a base64 audio file; spends credits by audio length. Risk: write

await corsair.elevenlabs.api.voiceChanger.convert({ voiceId: 'JBFqnCBsd6RMkjVDRZzb', fileBase64: '<base64 audio>', modelId: 'eleven_multilingual_sts_v2' });

Input

Name	Type	Required	Description
`voiceId`	`string`	Yes	Voice id, for example FGY2WhTYpPnrIDTdsKH5. Find ids with voices.list.
`fileBase64`	`string`	Yes	The file, base64-encoded (at most 100 MB once decoded).
`fileName`	`string`	No	—
`contentType`	`string`	No	Media type of the file, for example audio/mpeg.
`modelId`	`string`	No	Model id, for example eleven_multilingual_sts_v2. Default eleven_english_sts_v2.
`voiceSettings`	`object`	No	—
`seed`	`number`	No	—
`removeBackgroundNoise`	`boolean`	No	—
`fileFormat`	`pcm_s16le_16 \| other`	No	pcm_s16le_16 for raw 16-bit 16 kHz mono PCM, which is faster. Default other.
`outputFormat`	`string`	No	codec_samplerate_bitrate, for example mp3_22050_32 or mp3_44100_128 (the usual default). Higher quality formats may need a higher plan.
`enableLogging`	`boolean`	No	false turns on zero retention (enterprise plans): no history item, so the audio can only come back in this result.
`includeContent`	`boolean`	No	false leaves contentBase64 out of the result and returns its metadata only, which keeps a long clip out of an agent’s context. Read the audio later with history.getAudio. Default true.
`maxBytes`	`number`	No	Largest audio to carry in the result, in bytes. Over it, contentBase64 is left out and contentOmitted says so. Default 700000 (about 43 s of 128 kbps mp3), so the base64 fits the 1,000,000 byte cap on a script’s result on Corsair’s cloud. At most 50 MB.

voiceSettings full type

{
  stability?: number | null,
  similarity_boost?: number | null,
  style?: number | null,
  use_speaker_boost?: boolean | null,
  speed?: number | null
}

Output

Name	Type	Required	Description
`contentBase64`	`string`	No	The output, base64-encoded. Missing when contentOmitted is set.
`contentType`	`string`	Yes	Media type of the output, for example audio/mpeg.
`byteLength`	`number`	Yes	Size of the output in bytes.
`contentOmitted`	`not_requested \| over_limit`	No	Set when contentBase64 was left out: includeContent was false, or the output was over the limit.
`note`	`string`	No	How to get output that was left out.
`songId`	`string`	No	The `song-id` header, on generated music.
`requestId`	`string`	No	The `request-id` header. Pass it as previousRequestIds for continuity between clips.
`traceId`	`string`	No	The `x-trace-id` header; quote it to ElevenLabs support.
`characterCost`	`number`	No	Credits charged, from the `character-cost` header. Music and stem separation don’t send it; account.usageStats has their cost.
`historyItemId`	`string`	No	The history item holding this audio. Read it again with history.getAudio. Only speech operations have one.

Voice Design

create

voiceDesign.create Save a designed preview as a voice in this account; uses a voice slot. Risk: write

await corsair.elevenlabs.api.voiceDesign.create({ voiceName: 'Narrator', voiceDescription: 'A calm middle-aged narrator with a warm low voice.', generatedVoiceId: '<generated_voice_id>' });

Input

Name	Type	Required	Description
`voiceName`	`string`	Yes	—
`voiceDescription`	`string`	Yes	—
`generatedVoiceId`	`string`	Yes	generated_voice_id from a preview of voiceDesign.design or voiceDesign.remix.
`labels`	`Record<string, string>`	No	—

Needs a free voice slot; over voice_limit (see account.subscription) ElevenLabs answers 400. Output

Name	Type	Required	Description
`voice_id`	`string`	Yes	—
`name`	`string \| null`	No	—
`category`	`string \| null`	No	—
`description`	`string \| null`	No	—
`labels`	`Record<string, any> \| null`	No	—
`preview_url`	`string \| null`	No	—
`settings`	`object \| null`	No	—
`is_owner`	`boolean \| null`	No	—
`created_at_unix`	`number \| null`	No	—

settings full type

{
  stability?: number | null,
  similarity_boost?: number | null,
  style?: number | null,
  use_speaker_boost?: boolean | null,
  speed?: number | null
} | null

design

voiceDesign.design Generate voice previews from a description; spends credits. Hear one with voiceDesign.getPreview, then pass its generated_voice_id to voiceDesign.create. Risk: write

await corsair.elevenlabs.api.voiceDesign.design({ voiceDescription: 'A calm middle-aged narrator with a warm low voice.', autoGenerateText: true, outputFormat: 'mp3_22050_32' });

Input

Name	Type	Required	Description
`voiceDescription`	`string`	Yes	What the voice sounds like, 20 to 1000 characters.
`modelId`	`string`	No	eleven_multilingual_ttv_v2 (default) or eleven_ttv_v3.
`shouldEnhance`	`boolean`	No	—
`text`	`string`	No	What the previews say, 100 to 1000 characters. Or set autoGenerateText.
`autoGenerateText`	`boolean`	No	—
`loudness`	`number`	No	—
`seed`	`number`	No	—
`guidanceScale`	`number`	No	—
`outputFormat`	`string`	No	The format voiceDesign.getPreview will return, for example mp3_22050_32.

Cost follows the preview text, at about a credit a character in our tests (101 credits with a 101-character text, 200 with autoGenerateText). Audio doesn’t come back here. Previews are always made for streaming, which is what lets voiceDesign.getPreview read them, and duration_secs reads -1 for them. Output

Name	Type	Required	Description
`previews`	`object[]`	Yes	—
`text`	`string \| null`	No	—
`note`	`string`	No	—

previews full type

{
  generated_voice_id: string,
  media_type?: string | null,
  duration_secs?: number | null,
  language?: string | null
}[]

getPreview

voiceDesign.getPreview Read a preview’s audio by its generated_voice_id, without generating again. Subject to a size limit you can raise with maxBytes. Risk: read

await corsair.elevenlabs.api.voiceDesign.getPreview({ generatedVoiceId: '<generated_voice_id>', maxBytes: 2000000 });

Input

Name	Type	Required	Description
`generatedVoiceId`	`string`	Yes	generated_voice_id from a preview of voiceDesign.design or voiceDesign.remix.
`includeContent`	`boolean`	No	false returns the size and type only. Default true.
`maxBytes`	`number`	No	Largest audio to carry in the result, in bytes. Over it, contentBase64 is left out and contentOmitted says so. Default 700000 (about 43 s of 128 kbps mp3), so the base64 fits the 1,000,000 byte cap on a script’s result on Corsair’s cloud. At most 50 MB.

Reading a preview spent no credits in our test, read twice. outputFormat on the design or remix sets its format. We didn’t test how long a preview stays readable. Output

Name	Type	Required	Description
`contentBase64`	`string`	No	The output, base64-encoded. Missing when contentOmitted is set.
`contentType`	`string`	Yes	Media type of the output, for example audio/mpeg.
`byteLength`	`number`	Yes	Size of the output in bytes.
`contentOmitted`	`not_requested \| over_limit`	No	Set when contentBase64 was left out: includeContent was false, or the output was over the limit.
`note`	`string`	No	How to get output that was left out.
`songId`	`string`	No	The `song-id` header, on generated music.
`requestId`	`string`	No	The `request-id` header. Pass it as previousRequestIds for continuity between clips.
`traceId`	`string`	No	The `x-trace-id` header; quote it to ElevenLabs support.
`characterCost`	`number`	No	Credits charged, from the `character-cost` header. Music and stem separation don’t send it; account.usageStats has their cost.
`historyItemId`	`string`	No	The history item holding this audio. Read it again with history.getAudio. Only speech operations have one.

remix

voiceDesign.remix Generate previews of a changed version of an existing voice; spends credits. Hear one with voiceDesign.getPreview. Risk: write

await corsair.elevenlabs.api.voiceDesign.remix({ voiceId: '<your voice id>', voiceDescription: 'A little brighter.', autoGenerateText: true });

Input

Name	Type	Required	Description
`voiceId`	`string`	Yes	Voice id, for example FGY2WhTYpPnrIDTdsKH5. Find ids with voices.list.
`voiceDescription`	`string`	Yes	The changes to make to the voice.
`promptStrength`	`number`	No	—
`text`	`string`	No	What the previews say, 100 to 1000 characters. Or set autoGenerateText.
`autoGenerateText`	`boolean`	No	—
`loudness`	`number`	No	—
`seed`	`number`	No	—
`guidanceScale`	`number`	No	—
`outputFormat`	`string`	No	The format voiceDesign.getPreview will return, for example mp3_22050_32.

Only works on your own custom voices. It makes previews, read with voiceDesign.getPreview, and leaves the source voice as it is. In our tests it cost 101 credits with a 101-character text and 350 with autoGenerateText. Output

Name	Type	Required	Description
`previews`	`object[]`	Yes	—
`text`	`string \| null`	No	—
`note`	`string`	No	—

previews full type

{
  generated_voice_id: string,
  media_type?: string | null,
  duration_secs?: number | null,
  language?: string | null
}[]

Voices

addFromLibrary

voices.addFromLibrary Add a voice from the shared library to this account; uses a voice slot. Risk: write

await corsair.elevenlabs.api.voices.addFromLibrary({ publicUserId: '<public_owner_id>', voiceId: '<voice id>', newName: 'Narrator' });

Input

Name	Type	Required	Description
`publicUserId`	`string`	Yes	public_owner_id from voices.searchLibrary.
`voiceId`	`string`	Yes	Voice id, for example FGY2WhTYpPnrIDTdsKH5. Find ids with voices.list.
`newName`	`string`	Yes	The name it gets in your voices.

Output

Name	Type	Required	Description
`voice_id`	`string`	Yes	—

defaultSettings

voices.defaultSettings The settings a voice has by default. Risk: read

await corsair.elevenlabs.api.voices.defaultSettings({});

Input: empty object Output

Name	Type	Required	Description
`stability`	`number \| null`	No	—
`similarity_boost`	`number \| null`	No	—
`style`	`number \| null`	No	—
`use_speaker_boost`	`boolean \| null`	No	—
`speed`	`number \| null`	No	—

delete

voices.delete Delete a voice from this account, for good. Risk: destructive · Irreversible

await corsair.elevenlabs.api.voices.delete({ voiceId: '<voice id>' });

Input

Name	Type	Required	Description
`voiceId`	`string`	Yes	Voice id, for example FGY2WhTYpPnrIDTdsKH5. Find ids with voices.list.

Works on your own voices and library copies. It can’t be undone. Output

Name	Type	Required	Description
`status`	`string \| null`	No	—

edit

voices.edit Change a voice’s name, description or labels. Risk: write

await corsair.elevenlabs.api.voices.edit({ voiceId: '<voice id>', name: 'Narrator', description: 'For product videos' });

Input

Name	Type	Required	Description
`voiceId`	`string`	Yes	Voice id, for example FGY2WhTYpPnrIDTdsKH5. Find ids with voices.list.
`name`	`string`	Yes	The voice’s name; the API requires it.
`description`	`string`	No	—
`labels`	`Record<string, string>`	No	—

Output

Name	Type	Required	Description
`status`	`string \| null`	No	—

editSettings

voices.editSettings Change a voice’s stored settings (stability, similarity and so on). Risk: write

await corsair.elevenlabs.api.voices.editSettings({ voiceId: '<voice id>', settings: { stability: 0.4, similarity_boost: 0.7 } });

Input

Name	Type	Required	Description
`voiceId`	`string`	Yes	Voice id, for example FGY2WhTYpPnrIDTdsKH5. Find ids with voices.list.
`settings`	`object`	Yes	The settings to store, with the API’s field names.

settings full type

{
  stability?: number | null,
  similarity_boost?: number | null,
  style?: number | null,
  use_speaker_boost?: boolean | null,
  speed?: number | null
}

Output

Name	Type	Required	Description
`status`	`string \| null`	No	—

get

voices.get Read one voice with its settings and labels. Risk: read

await corsair.elevenlabs.api.voices.get({ voiceId: 'FGY2WhTYpPnrIDTdsKH5' });

Input

Name	Type	Required	Description
`voiceId`	`string`	Yes	Voice id, for example FGY2WhTYpPnrIDTdsKH5. Find ids with voices.list.

Output

Name	Type	Required	Description
`voice_id`	`string`	Yes	—
`name`	`string \| null`	No	—
`category`	`string \| null`	No	—
`description`	`string \| null`	No	—
`labels`	`Record<string, any> \| null`	No	—
`preview_url`	`string \| null`	No	—
`settings`	`object \| null`	No	—
`is_owner`	`boolean \| null`	No	—
`created_at_unix`	`number \| null`	No	—

settings full type

{
  stability?: number | null,
  similarity_boost?: number | null,
  style?: number | null,
  use_speaker_boost?: boolean | null,
  speed?: number | null
} | null

getSettings

voices.getSettings Read a voice’s stored settings. Risk: read

await corsair.elevenlabs.api.voices.getSettings({ voiceId: 'FGY2WhTYpPnrIDTdsKH5' });

Input

Name	Type	Required	Description
`voiceId`	`string`	Yes	Voice id, for example FGY2WhTYpPnrIDTdsKH5. Find ids with voices.list.

Output

Name	Type	Required	Description
`stability`	`number \| null`	No	—
`similarity_boost`	`number \| null`	No	—
`style`	`number \| null`	No	—
`use_speaker_boost`	`boolean \| null`	No	—
`speed`	`number \| null`	No	—

list

voices.list List the voices this account can use. Risk: read

await corsair.elevenlabs.api.voices.list({ search: 'narrator', pageSize: 10 });

Input

Name	Type	Required	Description
`nextPageToken`	`string`	No	next_page_token from the previous page.
`pageSize`	`number`	No	1 to 100. Default 10.
`search`	`string`	No	—
`sort`	`created_at_unix \| name`	No	—
`sortDirection`	`asc \| desc`	No	—
`voiceType`	`string`	No	personal, community, default, workspace, non-default or saved.
`category`	`string`	No	premade, cloned, generated or professional.
`voiceIds`	`string[]`	No	—

Output

Name	Type	Required	Description
`voices`	`object[]`	Yes	—
`has_more`	`boolean \| null`	No	—
`total_count`	`number \| null`	No	—
`next_page_token`	`string \| null`	No	—

voices full type

{
  voice_id: string,
  name?: string | null,
  category?: string | null,
  description?: string | null,
  labels?: Record<string, any> | null,
  preview_url?: string | null,
  settings?: {
    stability?: number | null,
    similarity_boost?: number | null,
    style?: number | null,
    use_speaker_boost?: boolean | null,
    speed?: number | null
  } | null,
  is_owner?: boolean | null,
  created_at_unix?: number | null
}[]

searchLibrary

voices.searchLibrary Search the shared voice library. Risk: read

await corsair.elevenlabs.api.voices.searchLibrary({ search: 'calm narrator', language: 'en', pageSize: 10 });

Input

Name	Type	Required	Description
`search`	`string`	No	—
`page`	`number`	No	Page number from 0.
`pageSize`	`number`	No	1 to 100. Default 30.
`category`	`string`	No	—
`gender`	`string`	No	—
`age`	`string`	No	—
`accent`	`string`	No	—
`language`	`string`	No	—
`locale`	`string`	No	—
`useCases`	`string[]`	No	—
`descriptives`	`string[]`	No	—
`featured`	`boolean`	No	—
`sort`	`created_date \| usage_character_count_1y \| trending \| cloned_by_count`	No	—

Output

Name	Type	Required	Description
`voices`	`object[]`	Yes	—
`has_more`	`boolean \| null`	No	—
`total_count`	`number \| null`	No	—

voices full type

{
  public_owner_id?: string | null,
  voice_id: string,
  name?: string | null,
  accent?: string | null,
  gender?: string | null,
  age?: string | null,
  category?: string | null,
  language?: string | null,
  description?: string | null,
  preview_url?: string | null
}[]

​Account

​subscription

​usageStats

​user

​Agents

​conversationDelete

​conversationGet

​conversationsList

​create

​delete

​get

​list

​signedUrl

​update

​Audio Isolation

​isolate

​Dubbing

​create

​delete

​get

​getAudio

​getTranscript

​list

​Forced Alignment

​create

​History

​delete

​get

​getAudio

​list

​Models

​list

​Music

​compose

​plan

​separateStems

​Single Use Tokens

​create

​Sound Effects

​generate

​Speech To Text

​convert

​deleteTranscript

​getTranscript

​Text To Dialogue

​convert

​Text To Speech

​convert

​convertWithTimestamps

​Voice Changer

​convert

​Voice Design

​create

​design

​getPreview

​remix

​Voices

​addFromLibrary

​defaultSettings

​delete

​edit

​editSettings

​get

​getSettings

​list

​searchLibrary

Account

subscription

usageStats

user

Agents

conversationDelete

conversationGet

conversationsList

create

delete

get

list

signedUrl

update

Audio Isolation

isolate

Dubbing

create

delete

get

getAudio

getTranscript

list

Forced Alignment

create

History

delete

get

getAudio

list

Models

list

Music

compose

plan

separateStems

Single Use Tokens

create

Sound Effects

generate

Speech To Text

convert

deleteTranscript

getTranscript

Text To Dialogue

convert

Text To Speech

convert

convertWithTimestamps

Voice Changer

convert

Voice Design

create

design

getPreview

remix

Voices

addFromLibrary

defaultSettings

delete

edit

editSettings

get

getSettings

list

searchLibrary