ElevenLabs Text to Speech (v3, v4, Multilingual v2)

What this Actor does

It turns text into speech with ElevenLabs models and gives you an MP3 file for each text. You can use any public voice from the ElevenLabs Voice Library, in any language the model supports.

Each text needs at least 300 characters. Shorter texts are rejected before the run starts, and you are not charged for them. Join short lines into one text. Every text has a fixed start-up time of about 15 seconds, so a few longer texts finish faster than many very short ones.

A single text can be up to 250,000 characters, which is enough for a long audiobook chapter or a full video script. A long text still comes back as one MP3, so you don't have to cut it up yourself.

On the Apify Free plan the Actor works as a one-time trial of 1,000 characters in total. Any paid Apify plan removes this limit.

Contact and partnerships

Questions, a problem with a run, higher volume, or an idea for working together? Message us on Telegram or email support@speakvela.com. If it is about a specific run, include the run ID so we can find it quickly.

Models and pricing

You pay per character of text, with no rounding. Every model has the same price, including Eleven v4 and v3.

Model	Price per 1,000 characters	Price per 1 million characters
Eleven v4	$0.01	$10
Eleven v4 Turbo	$0.01	$10
Eleven v3	$0.01	$10
Multilingual v2 (no Vietnamese)	$0.01	$10
Flash v2.5	$0.01	$10
Turbo v2.5	$0.01	$10

Each run also has a $0.001 start fee.

• You are charged only after the MP3 has been saved to your run's storage.
• Failed texts and texts refused by the content policy are not charged.
• Every voice costs the same.
• If a text would go over your "Maximum cost per run", the Actor skips it and tells you, instead of charging part of it.

On the Apify Free plan you can try the Actor once, with up to 1,000 characters in total, one text at a time. Any paid Apify plan removes these limits.

How to use it

1. Paste your text (at least 300 characters) into Text. Add more texts under More texts if you want several files from one run.
2. Set Voice ID. Open a voice in the ElevenLabs Voice Library and copy its ID, or keep the default voice.
3. Pick a Model. Eleven v3 and v4 sound the most natural and handle emotion well. Flash v2.5 is the fastest. Multilingual v2 is a good choice for long narration. Note: Multilingual v2 does NOT support Vietnamese.
4. Click Start. A text of a few hundred characters usually finishes in under a minute.

Output

Each text gives one MP3 in the run's key-value store (audio.mp3, or audio-001.mp3, audio-002.mp3… for several texts) and one row in the dataset:

{
  "index": 1,
  "status": "completed",
  "audioUrl": "https://api.apify.com/v2/key-value-stores/…/records/audio.mp3",
  "durationSec": 24.8,
  "characters": 349,
  "billedCharacters": 349,
  "modelId": "eleven_multilingual_v2",
  "voiceId": "auq43ws1oslv0tO4BDa7",
  "voiceName": "Adam Stone - Smooth and Relaxed",
  "text": "Welcome to our channel! Today we are going to talk about the history of coffee. Long before it reached cafes around the world, …"
}

A failed text has "status": "failed", an error message and "billedCharacters": 0. A text under 300 characters (after trimming spaces) has errorCode text_too_short. When the content policy refuses part of a text, errorCode is content_blocked and blockedText shows which part, so you can edit it and run again.

Input settings

Setting	Default	Notes
text		300 to 250,000 characters
texts		up to 100 extra texts per run, each 300+ characters
voiceId	auq43ws1oslv0tO4BDa7	any public ElevenLabs Voice Library voice
modelId	eleven_multilingual_v2	see the pricing table
language	auto	ISO code such as en, es, de, ja
stability	0.5	0 to 1
similarity	0.75	0 to 1
style		0 to 1, Multilingual v2 only
speed	1	0.7 to 1.2
speakerBoost	true

Use it from code

Run the Actor through the Apify API and wait for the result:

curl -X POST "https://api.apify.com/v2/acts/speakvela~elevenlabs-text-to-speech/run-sync-get-dataset-items?token=YOUR_APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{ "text": "Hello from the Apify API. This request turns a short article into speech with Eleven v3. Each text sent to this Actor needs at least 300 characters, so an API call usually carries a full paragraph, a product description or a scene from a video script rather than a single line. The MP3 is saved in the run storage.", "voiceId": "auq43ws1oslv0tO4BDa7", "modelId": "eleven_v3" }'

The response is the dataset, with an audioUrl for each text. The Apify JavaScript and Python clients, Make, Zapier and n8n integrations work the same way.

FAQ

Can I use my cloned or private voices? No. Only public voices from the ElevenLabs Voice Library are supported.

How long are files kept? As long as Apify keeps your run storage (see your Apify plan's data retention).

Can I use the audio commercially? The Actor returns the audio generated by ElevenLabs models. Check the ElevenLabs terms and the terms of the voice you pick.

I found a problem or need higher volume. Message us on Telegram or write to support@speakvela.com.