Introduction
Translate a podcast, voice note, narration, or audio advertisement by supplying audio_url and a target language. Set the source language and speaker count when known, and choose whether to retain music and ambience. The result is MP3 audio for listening and review.
| Field | Value |
|---|---|
| Input | Audio |
| Output | Audio |
| Publisher and product | ElevenLabs — ElevenLabs Dubbing |
| Release date | October 10, 2023 |
| Public audio API name | elevenlabs/dubbing/audio-to-audio |
| Public video API name | elevenlabs/dubbing/video-translate |
| Source duration limit | 15 minutes; longer sources rejected at submission |
| Target languages | 33; |
| Source languages | 57 plus auto detection |
| Speaker count | 1–32; omit |
| Background sound |
|
| Audio output observed | MP3; 4.50 s, 44.1 kHz mono, 128 kbps in our audio test |
| Video output observed | MP4; 6.04 s with new mono AAC audio and byte-identical video stream in our test |
| Billing unit | Seconds of source media, rounded up |
Key features & highlights
An MP3 dub or a translated MP4
Audio-to-audio accepts a recording and returns dubbed MP3 audio. Video-translate returns an MP4 with the picture preserved and its audio track replaced. Both accept sources up to 15 minutes, so choose the endpoint according to the media you need back.
33 targets, with speaker detection
Choose from 33 target languages and 57 source languages, or let source detection run automatically. Set speaker_count between 1 and 32 when the cast is known; omit it for detection. ElevenLabs says its dubbing carries each speaker’s voice character into the translated speech.
Retain the recording’s background
Enable keep_background_audio when music or ambience belongs in the translated recording. Review the balance with the new voices before using the result. For video, this soundtrack decision does not alter the filmed image or produce subtitles.
9.6 seconds for our audio sample
Our 4.5-second English-to-Spanish audio test returned a 4.50-second MP3 after about 9.6 seconds of processing. The file was 44.1 kHz mono at 128 kbps. A separate 6.04-second video took about 14.8 seconds and returned MP4 with a new mono AAC track; these are sample observations.
Request parameters
| Parameter | Required | Description |
|---|---|---|
audio_url | Yes | Provide the recording to translate through |
target_language | No | Select the destination language from 33 supported choices; en is the default. Set this explicitly when you need Spanish or another language rather than English. |
source_language | No | Choose auto for detection or identify the spoken language from the 57 supported source languages. |
speaker_count | No | Supply an integer from 1 to 32 for the distinct speakers, or omit this field for detection. Count people rather than dialogue turns when preparing an interview. |
keep_background_audio | No | Enable this option to keep music and ambience beneath the translated speech. Review that retained background with the dubbed voices before delivering the audio. Default: |
Pricing
$0.012 / second
Billing follows the source recording’s duration in seconds, rounded up to the next whole second, with sources up to 15 minutes. Language and speaker choices do not change the rate, and failed tasks carry no charge.
Use cases
A podcast interview in Spanish
Submit a 12-minute English interview through audio_url, choose es, and set speaker_count to 2 for its host and guest. The audio endpoint returns a translated MP3 for review and inclusion in a Spanish edition.
A spoken customer update in English
Provide a 45-second voice note recorded by a colleague who agreed to translation, with en as the target. Receive dubbed MP3 speech that you can review before sharing the update.
A narrated lesson with its visuals
Send a 6-minute screen-recorded lesson through video_url and select the intended target language. The returned MP4 keeps the screen demonstration while replacing the spoken track; create captions separately if needed.
An audio advertisement with music
Use a 30-second recorded advertisement you have rights to use and enable keep_background_audio. The output is an MP3 with translated voices and the retained music or ambience beneath them.
A product demonstration for another market
Supply a 90-second product walkthrough as video input and choose a supported target language. Receive an MP4 with the same picture and a dubbed soundtrack, suitable for a language review before release.
Prompt guide & pro tips
Prepare a permitted recording
Choose an audio or video source no longer than 15 minutes, using your own recording or speakers who agreed to dubbing. Confirm that you have the rights to the speech and any music.
Choose languages and speakers
Select one of 33 target languages and use automatic detection or one of 57 source languages. Enter
speaker_countfrom 1 to 32 when known, and decide whether to retain the background soundtrack.Submit the dubbing request
POST https://api.spicyapi.ai/api/v1/jobs/createTask with Bearer authentication and JSON such as {"model":"elevenlabs/dubbing/audio-to-audio","input":{"
audio_url":"https://example.com/narration.mp3","target_language":"es"}}. For video, select elevenlabs/dubbing/video-translate and supplyvideo_urlinstead.Retrieve and review the media
Read the task with GET /api/v1/jobs/recordInfo?taskId= until its state is succeeded, then download the returned asset. Check the translated speech and speaker changes; a video dub keeps the filmed lip movements.
ElevenLabs Dubbing API for audio translation
ElevenLabs Dubbing API translates an audio recording into dubbed MP3 speech in one of 33 target languages. This endpoint is for podcasts, voice notes, narration, and audio advertisements with source recordings up to 15 minutes.
ElevenLabs says in its dubbing documentation that dubbing preserves the voice characteristics of the original speakers. Review that resemblance on your own material: our short test establishes an output format and processing observation, rather than a broad assessment of translation or voice quality.
Submit an audio dubbing task
Send the public model name elevenlabs/dubbing/audio-to-audio with the recording and language settings inside input. POST the following JSON to https://api.spicyapi.ai/api/v1/jobs/createTask, using Authorization: Bearer <SPICY_API_KEY> and Content-Type: application/json.
{
"model": "elevenlabs/dubbing/audio-to-audio",
"input": {
"audio_url": "https://example.com/english-interview.mp3",
"target_language": "es",
"source_language": "en",
"speaker_count": 2,
"keep_background_audio": true
}
}
This illustrative request asks for Spanish speech from an English interview with two speakers and retained background sound. Replace the example URL with your recording; omit speaker_count when you want the service to detect the speakers.
Read the completed MP3 result
The completed task supplies a media asset containing the dubbed MP3. Use the returned taskId with GET /api/v1/jobs/recordInfo?taskId=<taskId> and the same authentication, then read the task state before downloading the asset.
{
"code": 200,
"data": {
"taskId": "<taskId>",
"state": "succeeded",
"output": {
"assets": [
{
"url": "https://example.com/dubbed-interview.mp3",
"mime": "audio/mpeg"
}
]
}
}
}
This trimmed response illustrates the completed-task shape; its URL is a placeholder. A submitted task is not a finished recording, so wait for succeeded before treating the asset as ready.
Audio parameters and limits
The audio request combines a required recording with language, speaker, and background choices.
| Parameter | Accepted value or scope | Practical use |
|---|---|---|
audio_url | Source audio, up to 15 minutes | Supply the recording you have rights to translate |
target_language | 33 languages; default en | Set es for Spanish output |
source_language | auto or 57 source languages | Identify a known source language or use detection |
speaker_count | 1–32; omit for detection | Count distinct people in an interview |
keep_background_audio | Background retention switch | Keep music and ambience under dubbed voices |
Recordings beyond 15 minutes are rejected at submission. Split a long episode at sensible topic or speaker boundaries, and review the joins when arranging the translated parts.
Pricing by source duration
Audio dubbing is billed per second of the source recording, rounded up to a whole second. Refer to the price shown on this page when budgeting a recording; the processing wait is not the billed media duration.
The submitted source length is the basis for the calculation, rather than the MP3’s file size. Prepare a short excerpt when assessing an unfamiliar language or a difficult speaker transition before submitting a longer recording.
Our English-to-Spanish audio test
Our 4.5-second English source returned a 4.50-second Spanish MP3 after about 9.6 seconds of processing. The output used 44.1 kHz sampling, one audio channel, and 128 kbps encoding.
The matching duration describes this sample and does not establish identical timing for every translation. These observations also do not establish a processing time for a 15-minute podcast or a result across all 33 target languages.
Improve the recording and review the dub
Clear dialogue and an explicit review plan help you judge whether the translated recording is suitable for its audience. Use a representative excerpt containing the names, terminology, or speaker changes that matter to the complete project.
- Select a passage with understandable speech and confirm permission from its speakers.
- Set the known source language, then choose the intended target language explicitly.
- Enter the distinct speaker count when known, or leave detection enabled by omitting it.
- Decide whether background music belongs in the delivery and set retention accordingly.
- Have a listener familiar with the target language check meaning, pronunciation, and turn changes.
A clean narration sample does not represent overlapping interview dialogue. Listen through interruptions and compare who speaks each translated line; a numeric speaker count is not a substitute for that review.
Troubleshoot or choose another workflow
Correct an overlong source or an out-of-range speaker count before resubmitting a rejected request. If a task ends unsuccessfully, read its error details and check the submitted recording and language choices before trying again.
Choose elevenlabs/dubbing/video-translate when you need an MP4 with translated speech over its source picture. For video that also needs lip movement, consider HeyGen Video Translate; Video Translator V1 is another video translation option. Use your own recordings or speakers who agreed, and never impersonate a real person.
API quick start
# Generate the idempotency key once; reuse it when retrying after a lost response.
export SPICY_IDEMPOTENCY_KEY="${SPICY_IDEMPOTENCY_KEY:-$(uuidgen)}"
curl -X POST https://api.spicyapi.ai/api/v1/jobs/createTask \
-H "Authorization: Bearer $SPICY_API_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: $SPICY_IDEMPOTENCY_KEY" \
-d '{
"model": "elevenlabs/dubbing/audio-to-audio",
"input": {
"audio_url": "https://example.com/reference.mp3",
"source_language": "auto",
"target_language": "en"
}
}'
# Copy data.taskId from the response into TASK_ID, then poll the status.
TASK_ID="TASK_ID_FROM_RESPONSE"
curl "https://api.spicyapi.ai/api/v1/jobs/recordInfo?taskId=$TASK_ID" \
-H "Authorization: Bearer $SPICY_API_KEY"
See the API reference for authentication, the complete schema, polling, and error handling.




