Customize your input with more control.
Customize your input with more control.
Customize your input with more control.
Your request will cost $0.1 per 1000 character.
Your request will cost $0.1 per 1000 character.
Generate lifelike, expressive speech directly through fal's serverless infrastructure using Eleven v3, ElevenLabs' most advanced speech synthesis model. Eleven v3 delivers natural, emotionally-aware audio with broad dynamic range, inline audio tag control, and support for 70+ languages — all available through a single fal endpoint: fal-ai/elevenlabs/tts/eleven-v3.
Eleven v3 is ElevenLabs' flagship text-to-speech model, designed for the most expressive and emotionally nuanced AI-generated speech available today. Unlike earlier models, Eleven v3 offers a broad dynamic range controlled through inline audio tags — descriptive markers like [whispers], [laughs], [excited], or [sad] that the model interprets to shape delivery, emotion, tempo, and tone.
Key capabilities of Eleven v3:
[laughs], [whispers], [sighs], [excited], [sarcastically], and many more)text-to-dialogue/eleven-v3 endpointEleven v3 is the clear choice when expressive delivery matters most. For ultra-low-latency conversational use cases, pair it with Flash or Turbo v2.5; for pure multilingual narration with classic stability, Multilingual v2 remains a solid alternative.
ElevenLabs is an AI audio research and deployment company founded in 2022 by Piotr Dąbkowski (ex-Google ML engineer) and Mati Staniszewski (ex-Palantir). Headquartered in London and New York, the company has become the dominant platform in voice AI — according to recent market analysis, ElevenLabs accounts for approximately 98% of observed mid-market voice AI spend and is the entry point for 95% of first-time voice AI customers.
The company's platform spans text-to-speech, speech-to-text (Scribe), voice cloning (Instant and Professional), voice design, dubbing, music generation, and real-time conversational AI agents. ElevenLabs raised $500M at an $11B valuation in early 2026 and crossed $330M in annual recurring revenue in 2025.
ElevenLabs is used by 41% of Fortune 500 companies and a diverse set of industry leaders:
Enterprise customers choose ElevenLabs for SOC 2 Type II, ISO 27001, HIPAA, PCI DSS L1, and GDPR compliance, Zero Retention Mode, regional data residency (US, EU, India), and a library of 10,000+ voices.
fal brings ElevenLabs' Eleven v3 model into a single unified serverless AI platform alongside hundreds of other generative AI models — image, video, audio, and LLMs. Running Eleven v3 on fal gives you:
Pricing on fal is based on the length of the input text:
$0.10 per 1,000 characters
A 500-character paragraph costs $0.05. A 10,000-character short story costs $1.00. No seat licenses, no subscriptions, no minimums.
POST https://fal.run/fal-ai/elevenlabs/tts/eleven-v3
Model ID: fal-ai/elevenlabs/tts/eleven-v3
All requests require a fal API key. Set FAL_KEY as an environment variable in your runtime. Never expose your API key in client-side code — always proxy requests through a server you control.
bashexport FAL_KEY="your-fal-api-key"
| Parameter | Type | Required | Description |
|---|---|---|---|
text | string | ✅ | The text to convert to speech. Supports inline audio tags like [laughs], [whispers], [excited]. |
voice | string | — | Voice name or ID. Default: "Rachel". Examples: Aria, Roger, Sarah, Laura, Charlie, George, Callum, River, Liam, Charlotte, Alice, Matilda, Will, Jessica, Eric, Chris, Brian, Daniel, Lily, Bill. |
stability | float | — | Voice stability, 0–1. Lower values = more expressive variation, higher = more consistent delivery. Default: 0.5. |
similarity_boost | float | — | How closely the output matches the reference voice, 0–1. Default: 0.75. |
speed | float | — | Playback speed multiplier. Default: 1. |
language_code | string | — | ISO 639-1 language code to enforce a specific language. |
apply_text_normalization | enum | — | "auto", "on", or "off". Controls spelling out of numbers, abbreviations, etc. Default: "auto". |
seed | int | — | Random seed for reproducibility. |
timestamps | bool | — | When true, returns per-word timestamps in the response. |
output_format | enum | — | Codec, sample rate, and bitrate. Default: "mp3_44100_128". |
mp3_22050_32, mp3_44100_32, mp3_44100_64, mp3_44100_96, mp3_44100_128 (default), mp3_44100_192, pcm_8000, pcm_16000, pcm_22050, pcm_24000, pcm_44100, pcm_48000, ulaw_8000, alaw_8000, opus_48000_32, opus_48000_64, opus_48000_96, opus_48000_128, opus_48000_192.
| Field | Type | Description |
|---|---|---|
audio | File | The generated audio file, returned as { "url": "https://v3.fal.media/..." }. |
timestamps | array | Per-word timestamps. Only returned when timestamps: true in the request. |
json{ "text": "Hello! [excited] This is a test of the text to speech system, powered by ElevenLabs. [whispers] How does it sound?", "voice": "Aria", "stability": 0.5, "similarity_boost": 0.75, "speed": 1, "apply_text_normalization": "auto" }
json{ "audio": { "url": "https://v3.fal.media/files/zebra/zJL_oRY8h5RWwjoK1w7tx_output.mp3" } }
Install the client:
bashnpm install --save @fal-ai/client
Basic synchronous call:
javascriptimport { fal } from "@fal-ai/client"; const result = await fal.subscribe("fal-ai/elevenlabs/tts/eleven-v3", { input: { text: "Hello! This is a test of the text to speech system, powered by ElevenLabs. How does it sound?", voice: "Aria", stability: 0.5 }, logs: true, onQueueUpdate: (update) => { if (update.status === "IN_PROGRESS") { update.logs.map((log) => log.message).forEach(console.log); } }, }); console.log(result.data.audio.url); console.log(result.requestId);
Install the client:
bashpip install fal-client
pythonimport fal_client def on_queue_update(update): if isinstance(update, fal_client.InProgress): for log in update.logs: print(log["message"]) result = fal_client.subscribe( "fal-ai/elevenlabs/tts/eleven-v3", arguments={ "text": "Hello! This is a test of the text to speech system, powered by ElevenLabs. How does it sound?", "voice": "Aria", "stability": 0.5, }, with_logs=True, on_queue_update=on_queue_update, ) print(result["audio"]["url"])
bashcurl --request POST \ --url https://fal.run/fal-ai/elevenlabs/tts/eleven-v3 \ --header "Authorization: Key $FAL_KEY" \ --header "Content-Type: application/json" \ --data '{ "text": "Hello! This is a test of the text to speech system, powered by ElevenLabs. How does it sound?", "voice": "Aria" }'
Eleven v3 on fal supports streaming for real-time audio playback:
javascriptimport { fal } from "@fal-ai/client"; const stream = await fal.stream("fal-ai/elevenlabs/tts/eleven-v3", { input: { text: "Hello! This is a test of the text to speech system, powered by ElevenLabs." } }); for await (const event of stream) { console.log(event); } const result = await stream.done();
For long-form generation (audiobooks, podcasts, large batches), use the queue API with an optional webhook to receive completion notifications:
javascriptimport { fal } from "@fal-ai/client"; // Submit const { request_id } = await fal.queue.submit("fal-ai/elevenlabs/tts/eleven-v3", { input: { text: "Long-form narration goes here..." }, webhookUrl: "https://your-app.com/webhooks/tts-complete", }); // Check status later const status = await fal.queue.status("fal-ai/elevenlabs/tts/eleven-v3", { requestId: request_id, logs: true, }); // Fetch the result const result = await fal.queue.result("fal-ai/elevenlabs/tts/eleven-v3", { requestId: request_id });
Eleven v3's standout feature is inline audio tag control. Add bracketed directives anywhere in your text to shape how it's performed:
Emotional tags
[happy] [sad] [excited] [angry] [sarcastically] [nervous] [confident]
Delivery tags
[whispers] [shouting] [slowly] [quickly] [softly]
Non-verbal sounds
[laughs] [chuckles] [sighs] [gasps] [coughs] [gulps] [applause]
Accent tags
[strong canadian accent] [british accent] [southern accent]
Example:
text[slowly] Back then... [chuckles] we had no phones. [whispers] Just dirt roads and [coughs] big dreams. [sad] Then it happened.
Audio tags are voice- and context-dependent — some voices interpret certain tags more reliably than others. Experiment to find what works best for your use case.
Eleven v3's expressive range and multilingual reach make it well-suited for high-production-value audio work:
If Eleven v3 isn't the right fit, fal also hosts other ElevenLabs endpoints:
fal-ai/elevenlabs/text-to-dialogue/eleven-v3 — multi-speaker dialogue generation with matched prosodyfal-ai/elevenlabs/tts/multilingual-v2 — stable, high-quality narration across 29 languagesfal-ai/elevenlabs/tts/turbo-v2.5 — lowest-latency option, ideal for real-time agents