Skip to content

EL โ€‹

Updated

Endpoint: POST https://vod.intl.tencentcloudapi.comAction: CreateAigcAudioTask

EL (ElevenLabs) is an audio generation model. The compose_* versions generate music from a prompt, and the sound_t2s_* versions generate sound effects from text. Music versions compose_v2, compose_v2.5, and compose_detailed_v2.5, and sound effect versions sound_t2s_v2 and sound_t2s_v3 are available; compose_v2.5, compose_detailed_v2.5, and sound_t2s_v3 were launched on 2026-11-16.

See Version differences to pick a version and Calling by version for request examples.

Basic Information โ€‹

ItemValue
ModelNameEL
ModelVersionMusic: compose_v2 / compose_v2.5 / compose_detailed_v2.5; sound effects: sound_t2s_v2 / sound_t2s_v3
SceneTypeNot required โ€” leave it empty
Usecompose_* music, sound_t2s_* sound effects

Version differences โ€‹

ModelVersionTypeUpstream ElevenLabs APIUpstream model
compose_v2MusicCompose musicmusic_v2
compose_v2.5MusicCompose musicmusic_v2_5
compose_detailed_v2.5Music (detailed)Compose music with detailsmusic_v2_5
sound_t2s_v2Sound effectsSound effectsโ€”
sound_t2s_v3Sound effectsSound effectsโ€”

How to choose:

  • General music from a prompt โ†’ compose_v2.5 (the latest); compose_v2 remains available.
  • Music via the "compose with details" endpoint โ†’ compose_detailed_v2.5.
  • Short sound effects โ†’ sound_t2s_v3 (the latest); sound_t2s_v2 remains available.

The "Upstream model" column is inferred from the version names and the upstream model_id values (music_v2 / music_v2_5); the integration guide does not state the mapping. The upstream models of the two sound effect versions are not stated, so they are left empty; sound_t2s_v3 shares the same parameter table as v2 (Duration [1, 30]), and the guide does not describe the difference in output.

โš ๏ธ Upstream, "compose with details" additionally accepts a section-level composition_plan and returns metadata alongside the audio. The integration guide only documents Prompt + OutputConfig for compose_detailed_v2.5, and does not say whether composition_plan can be passed through AdditionalParameters, or whether the extra metadata is returned. Treat these as unconfirmed and verify with a test call before relying on them.

Request Parameters โ€‹

ParameterRequiredTypeDescription
SubAppIdOptionalIntegerVOD application ID
ModelNameOptionalStringEL
ModelVersionOptionalStringcompose_v2 / compose_v2.5 / compose_detailed_v2.5 / sound_t2s_v2 / sound_t2s_v3
SceneTypeOptionalStringNot required
PromptOptionalStringStyle / instrument / scene description (sound effects: describe the sound)
OutputConfigOptionalObjectOutput settings, see below

OutputConfig โ€‹

FieldApplies toDescription
StorageModeAll versionsTemporary: not stored in VOD; a temporary URL valid for 7 days is returned in the event notification. Permanent: stored in VOD; a FileId is returned
Durationsound_t2s_* onlySound effect length in seconds, [1, 30]

Calling by version โ€‹

Music (compose_v2.5) โ€‹

Same call for compose_v2 โ€” just change ModelVersion.

json
{
  "ModelName": "EL",
  "ModelVersion": "compose_v2.5",
  "Prompt": "A quiet electronic instrumental for a night drive",
  "SubAppId": 123456789,
  "OutputConfig": {
    "StorageMode": "Temporary"
  }
}

Detailed music (compose_detailed_v2.5) โ€‹

json
{
  "ModelName": "EL",
  "ModelVersion": "compose_detailed_v2.5",
  "Prompt": "Cinematic orchestral trailer music, slow build to an epic climax",
  "SubAppId": 123456789,
  "OutputConfig": {
    "StorageMode": "Permanent"
  }
}

Sound Effects (sound_t2s_v3) โ€‹

Same call for sound_t2s_v2 โ€” just change ModelVersion.

json
{
  "ModelName": "EL",
  "ModelVersion": "sound_t2s_v3",
  "Prompt": "Rain on a metal roof",
  "SubAppId": 123456789,
  "OutputConfig": {
    "StorageMode": "Temporary",
    "Duration": 6
  }
}

Response Example โ€‹

json
{
  "Response": {
    "TaskId": "<task-id>",
    "RequestId": "<request-id>"
  }
}

Query the completion result with DescribeTaskDetail. See Audio Result Structure.

Special Settings โ€‹

  • OutputConfig.Duration only applies to the sound effect versions sound_t2s_v2 / sound_t2s_v3 (1โ€“30 seconds). It is not applied to the compose_* music versions.
  • No version specifies SceneType.
  • Upstream, a plain prompt is limited to 4100 characters (reference value from the ElevenLabs API, not stated in the VOD guide).

Built with VitePress ยท Deployed on Tencent EdgeOne Pages