Imagen

Cheapest Text to Speech API for Developers

Speech generation and voice cloning from $21/month. Self-hosted speech is $0.001 per second of audio, not per character, and the $149/month Open Source Unlimited plan has no limit on it.

Get API key See pricing

$0.001 per second of audio
48 languages
Unlimited on the $149 plan

Last updated Oct 2, 2026 · By ModelsLab Engineering

Cheapest text to speech API: the short answer

Per minute of speech, the cheapest APIs are Google Cloud and Amazon Polly standard voices at $4 per million characters, about $0.004 a minute. ModelsLab self-hosted speech is $0.001 per second, or $0.06 a minute, which costs more per minute than Google, Polly, OpenAI tts-1 and ElevenLabs Flash.

ModelsLab is cheaper at high volume. On the $149/month Open Source Unlimited plan self-hosted speech has no per-second charge, so it costs less than ElevenLabs Flash above about 55 hours of speech a month, less than OpenAI tts-1 above about 184 hours, and less than Google or Polly standard voices above about 690 hours. Below those volumes, pick the provider in the table.

Text to Speech API Prices per Minute

List prices from each provider's own pricing page. Per-minute figures assume about 900 characters per minute of speech.

Provider	Voices	List price	About per minute	Unlimited plan breaks even at
Google Cloud	Standard voices	$4 per 1M characters	$0.0036	About 690 hours a month
Amazon Polly	Standard voices	$4 per 1M characters	$0.0036	About 690 hours a month
OpenAI	tts-1	$15 per 1M characters	$0.0135	About 184 hours a month
ElevenLabs	Flash / Turbo	$0.05 per 1K characters	$0.045	About 55 hours a month
ModelsLab	Self-hosted TTS and voice cloning	$0.001 per second of audio	$0.06	No per-second charge on the $149 plan

Prices checked September 27, 2026 on cloud.google.com/text-to-speech/pricing, aws.amazon.com/polly/pricing, developers.openai.com/api/docs/pricing and elevenlabs.io/pricing/api. The same figures are on the Best Text to Speech API 2026 page.

Where ModelsLab Is Cheaper, and Where It Is Not

Per-Second Pricing, Not Per-Character

Per minute, ModelsLab is not the cheapest text to speech API: Google Cloud and Amazon Polly standard voices cost about $0.004 a minute against $0.06 here. What changes the comparison is volume and what else you need from the same plan.

Most text-to-speech APIs bill per character. That model is fine for short UI prompts and punishing for anything long — an audiobook chapter, a training module, a dubbed video — because the bill scales with every word you generate, including the takes you throw away.

ModelsLab bills self-hosted speech by the audio it produces instead: $0.001 per second, with a $0.0047 minimum per request, paid from the usage included in your plan. $21 on Basic covers about 5.8 hours of speech and $47 on Standard about 13 hours. On the $149/month Open Source Unlimited plan there is no per-second charge at all, so retakes and long scripts cost nothing extra.

Cheap Does Not Mean a Thin Voice Catalogue

The low tier is not a stripped-down trial. Every plan reaches the same speech stack:

Voice cloning from a short reference sample
Text-to-speech across 48 languages
Voice-to-voice conversion for changing an existing recording
Speech-to-text transcription on the same key
Self-hosted speech models on ModelsLab GPUs
Optional partner voices (ElevenLabs, Inworld)
One REST endpoint and one API key across all of it

Speech to Text: $0.0047 per Request

Transcription runs on Whisper large-v3 on our own GPUs and costs $0.0047 per request, whatever the length of the audio. It is unlimited on the $149 plan. See the Speech to Text API for word timestamps and a quick start.

What Each Plan Costs

One ModelsLab subscription covers speech, image, video and LLM APIs. Self-hosted speech is paid from the usage included in your plan at $0.001 per second of generated audio, with a $0.0047 minimum per request.

Plan	Price / Month	Self-hosted speech included	Parallel generations
Basic	$21	About 5.8 hours	5
Standard	$47	About 13 hours	10
Open Source Unlimited (best value)	$149	No limit	15

Plan prices and the $0.001 per second audio rate checked 2026-09-23. Open Source Unlimited covers the speech models ModelsLab self-hosts, including voice cloning. Partner voices (ElevenLabs, Inworld) use the usage included in Basic and Standard first, then wallet balance, and wallet balance on Open Source Unlimited.

Generate Speech in One Request

One endpoint for text-to-speech and voice cloning.

Text to speech (Python)

Python

1import requests
2
3response = requests.post(
4    "https://modelslab.com/api/v6/voice/text_to_speech",
5    headers={"Content-Type": "application/json"},
6    json={
7        "key": "YOUR_MODELSLAB_API_KEY",
8        "voice_id": "madison",
9        "language": "american english",
10        "prompt": "Your order has shipped and arrives on Thursday.",
11    },
12)
13
14# Billed by the seconds of audio returned ($0.001/s, $0.0047 minimum)
15print(response.json())

Change the voice of an existing recording (Python)

Python

1import requests
2
3response = requests.post(
4    "https://modelslab.com/api/v6/voice/voice_to_voice",
5    headers={"Content-Type": "application/json"},
6    json={
7        "key": "YOUR_MODELSLAB_API_KEY",
8        # the recording to convert
9        "init_audio": "https://example.com/original-recording.wav",
10        # a clean sample of the voice you want it to sound like
11        "target_audio": "https://example.com/target-voice.wav",
12    },
13)
14
15# Keeps the timing and delivery of the source, swaps the voice
16print(response.json())

Batch a script with JavaScript

JavaScript

1const lines = [
2  'Chapter one. The house on the hill.',
3  'It had been empty for as long as anyone could remember.',
4];
5
6// One request per line, billed by the seconds of audio each returns.
7const audio = await Promise.all(
8  lines.map(async (prompt) => {
9    const res = await fetch('https://modelslab.com/api/v6/voice/text_to_speech', {
10      method: 'POST',
11      headers: { 'Content-Type': 'application/json' },
12      body: JSON.stringify({
13        key: 'YOUR_MODELSLAB_API_KEY',
14        voice_id: 'madison',
15        language: 'american english',
16        prompt,
17      }),
18    });
19    return res.json();
20  }),
21);
22
23console.log(`${audio.length} clips generated`);

Ship Speech in 3 Steps

From signup to production audio.

STEP 01

Step 1: Sign Up

Create a ModelsLab account, subscribe to the $21/month Basic plan, and copy your API key from the dashboard. Every plan reaches the same voices, and the same key works for image, video and LLM APIs.

STEP 02

Step 2: Generate Speech

POST your text and a voice_id to the text-to-audio endpoint, or pass a reference sample to clone a voice. Each request is billed at $0.001 per second of audio it returns, from the usage included in your plan.

STEP 03

Step 3: Move to Unlimited at $149/mo

When your speech volume passes what Standard covers (about 13 hours a month), the $149/month Open Source Unlimited plan removes the limit. It covers unlimited generation on the open-weight speech models ModelsLab self-hosts.

Get API Key

Related speech guides

Best Text to Speech API 2026

Full comparison beyond price — voice quality, latency, and language coverage.

Cheapest AI Image API

The same cost breakdown for image generation, down to $0.0015 per image at 100,000 images a month.

Voice Cloning API

Clone a voice from a short sample and generate speech in 48 languages.

How the Unlimited Plan Works

ModelsLab self-hosts its speech models on its own GPUs, which is what makes flat pricing possible: once the infrastructure is running, the marginal cost of another generation is close to zero. The $149/month Open Source Unlimited plan removes the per-second charge on those models.

Partner voices are the exception and are not part of "unlimited". ElevenLabs and Inworld voices are billed by the provider rate: from the usage included in Basic or Standard first, then from wallet balance, and from wallet balance on Open Source Unlimited.

Which Plan Fits Your Volume

Up to about 5.8 hours of generated speech a month, Basic at $21 is the cheapest ModelsLab plan for it. Up to about 13 hours, Standard at $47 covers it. Above that, Open Source Unlimited at $149 has no limit on self-hosted speech models, and it removes the need to forecast volume at all.

Voice cloning costs the same as plain text-to-speech — there is no separate cloning tier or per-voice fee. A reference sample needs at least 4 seconds of clean speech; longer samples are trimmed to about 10 seconds.

What You Get From $21/Month

Key advantages that set us apart

Plans from $21/month, $0.001 per second of speech

Billed by seconds of audio, not per character

Unlimited open-weight speech at $149/month

Voice cloning included at no extra cost

48 languages on the same endpoint

Voice-to-voice conversion as well as text-to-speech

Speech-to-text transcription on the same API key

Self-hosted text to speech, voice cloning and transcription

Partner voices billed at their own rate, shown before you call

One API key across image, video, speech and LLM

Our Popular Use Cases

What teams build with low-cost speech generation:

On Open Source Unlimited there is no meter at all: generate chapter after chapter and re-record individual paragraphs without a per-character or per-second charge.

Your Data is Secure: GDPR Compliant AI Services

ModelsLab GDPR Compliance Certification Badge

GDPR Compliant

Pricing That's Perfect

Choose plan as per your needs, cancel anytime.

100% refund policy on monthly & yearly plans — cancel anytime

Contact Sales

Best Value

Open Source Unlimited

Mission-Critical

$149/month

100% refund policy · cancel anytime

Unlimited Open Source Models

100% refund policy

24x7 Support

15 parallel generations

Access to all APIs

Unlimited generations on all open-source models

For mission critical workloads

Add Team Members

Priority GPU Clusters

Standard

Production

$47/month

100% refund policy · cancel anytime

Moderate Traffic

100% refund policy

Priority Developer Support

10 concurrent API requests

For Production workloads

API access to all models

Prototype

Basic

Prototype

$21/month

100% refund policy · cancel anytime

Moderate Traffic

100% refund policy

Developer Support via Discord/Email

5 concurrent API requests

API access to all models

Shared GPU

Speech to Text API
Transcription with word-level timestamps.
Whisper API
Whisper large-v3 on our own GPUs.
Text to Speech API
48 languages, billed per second of audio.
Voice Cloning API
Clone a voice from a short sample.
Cheapest AI API
Image, video and speech prices side by side.
Unlimited AI API
Every self-hosted open-source model for $149/mo.

Get Expert Support in Seconds

We're Here to Help.

Want to know more? You can email us anytime at support@modelslab.com

View Docs

Per minute, Google Cloud and Amazon Polly standard voices are the cheapest at $4 per million characters, about $0.004 a minute. OpenAI tts-1 is about $0.014 a minute and ElevenLabs Flash about $0.045. ModelsLab self-hosted speech is $0.001 per second ($0.06 a minute) on plans from $21/month, so it costs more per minute. It becomes cheaper on the $149/month Open Source Unlimited plan, which has no per-second charge on self-hosted speech: above about 55 hours of speech a month against ElevenLabs Flash, 184 hours against OpenAI tts-1 and 690 hours against Google or Polly standard voices. Prices checked 2026-09-27.

No. Self-hosted speech is billed by the length of the audio it produces: $0.001 per second, with a $0.0047 minimum per request, paid from the usage included in your plan. The number of characters you send does not change the price, and the $149/month Open Source Unlimited plan has no limit on the self-hosted speech models.

Voice cloning is included on every plan at the same rate as plain text-to-speech, $0.001 per second of generated audio. There is no separate cloning tier; extra saved-voice uploads cost $1 per 10 from the wallet. A reference sample needs at least 4 seconds of clean speech, and longer samples are trimmed to about 10 seconds.

Unlimited generation on the speech models ModelsLab self-hosts: text to speech, voice cloning and Whisper large-v3 transcription. Partner voices such as ElevenLabs and Inworld are billed from your wallet on this plan.

ModelsLab is a paid service. Plans start at $21/month. Creating an account is free, but generating speech requires an active plan.

The text to speech API covers 48 languages, including Hindi, Bengali, Urdu, Tamil, Telugu, Malayalam, Punjabi, Marathi, Gujarati, Kannada, Persian, Arabic and Spanish alongside English. The same endpoint and plan cover all of them.

Yes. Start on Basic at $21/month and move up as volume grows. Standard ($47) covers about 13 hours of generated speech a month; above that, the $149/month Open Source Unlimited plan removes the limit on self-hosted speech models. You can move back down if volume falls. There is no lock-in.