Is this list ranked from best to worst?

No. Entries are alphabetical inside each group, and the directory makes no quality judgement. Our ranked picks are on the best-of pages, which are written separately.

Has anyone paid to be listed here?

No. Listings are not paid, and the links are plain outbound links marked nofollow. If sponsored placements are ever added, they will be labelled as sponsored and will not change the order of editorial entries.

Can I use audio from a free plan commercially?

Often not. Several free plans in this directory are personal use only, or say commercial use needs a paid plan. The entry for each tool records what its site states, but confirm on the tool's current terms before you publish or sell anything.

Can I clone a voice with these tools?

Several list it. Fish Audio says a clone can be made from a 15-second clip, Cartesia lists instant voice cloning from its Pro plan, ElevenLabs lists it from Starter, and Hume AI's pricing table lists voice cloning on its free plan. Among the open-weight models, Chatterbox and F5-TTS clone from a reference clip, while Kokoro-82M uses a fixed set of voices and does not clone. Only clone a voice when you have the speaker's consent.

Which of these can I run on my own computer?

The models in the open-weight group: Chatterbox, F5-TTS, Kokoro-82M and Piper. Their hardware needs and licences differ, so read each entry. F5-TTS publishes non-commercial weights.

Why are some tools marked 'see site' for free access?

The tool's site did not state its free access in a form we could record, or its pricing page did not load when we checked. We would rather say so than guess.

How can a tool be added or corrected?

Use the contact page. Send the official URL and what the tool does, and we will check the facts on its own site before adding or changing an entry.

AI voice generator directory: every tool, free tier marked

A neutral list of AI voice and text-to-speech tools and open-weight speech models, with the official link, free option and licence terms each site states.

14 tools
A to Z, not a ranking
Last updated: 2026-10-04

Our picks

The short ranked list from our best-of page. The complete list is below.

See the full ranking

1
ElevenLabs
The most realistic free text-to-speech and voice cloning
2
Microsoft Edge / Azure TTS (Read Aloud)
Unlimited, frictionless free narration
3
Google Cloud Text-to-Speech
High-quality TTS for developers on a free monthly quota
4
Murf AI
A free trial of a full voiceover studio

This is a catalogue, not a ranking. It lists the AI voice and text-to-speech tools we have read an official page for, split into hosted tools and models you run yourself, with the official link and what each site says about free access.

Our recommendations live on a separate page. For picks, see Best free AI voice generators: what you get in 2026. Use this directory to see what exists, or to check a name you keep running into.

Showing 14 of 14

Voice and text-to-speech

Hosted tools and cloud APIs that read a script aloud, clone a voice or change one. Free plans here are usually small, and most of them stop at personal use.

Cartesia
Text to speech, speech to text and voice agents on a credit system. The free plan includes 20,000 credits per month. Pro is $5 per month with 100,000 credits, a commercial use license and instant voice cloning. Startup is $49 (1.25M credits, adds professional voice cloning) and Scale is $299 (8M credits). The page lists text to speech at roughly 750 to 800 credits per minute of audio and voice agents at $0.06 per minute.
Free optionDetails →
Descript
A text-based audio and video editor with AI voices, so voice generation is one feature among many. The free plan includes 60 media minutes a month and 100 one-time AI credits. Paid plans start with Hobbyist (10 media hours and 400 AI credits a month, 1080p exports without a watermark), then Creator (30 hours, 800 credits, 4K) and Business (40 hours, 1,500 credits). The page shows two prices per plan depending on billing period, so check it for the current figure.
Free optionDetails →
ElevenLabs
In our picks, #1
Text to speech, voice cloning, sound effects, music and speech to text on one credit system. The free plan gives 10,000 credits per month and does not allow commercial use. Starter adds a commercial license and instant voice cloning (30,000 credits), and Creator adds professional voice cloning (121,000 credits). Higher tiers are Pro at $99 per month (600,000 credits) and Scale at $299 (1.8M credits). The page lists introductory first-month prices on Starter and Creator.
Free optionDetails →
Fish Audio
Text to speech and voice cloning, with a flagship model the site calls S2.1 Pro and a library the site says holds over 2 million voices. The free plan includes monthly generations for personal use only, and the FAQ says monetizing content or commercial use needs a paid plan. The site says a clone can be made from a 15-second clip and lists languages including English, Japanese, Korean, Chinese, French, German, Arabic and Spanish.
Free optionDetails →
Hume AI
Text to speech and speech to speech with its Octave and EVI voices. The free plan includes 10,000 characters per month (about 10 minutes of speech) and 5 minutes of EVI, and the pricing table lists voice cloning as unlimited. Starter is $3 per month with 30,000 characters, Creator is $7 and Pro is $70. The page says commercial licensing is available on every tier.
Free optionDetails →
Murf
In our picks, #4
Voiceovers, AI voice agents and video translation, with 200+ voices in 35+ languages. The free plan allows 10 minutes of voice generation, and commercial use needs a paid plan. The text-to-speech API is listed at $0.01 per minute.
Free optionDetails →
OpenAI text to speech
Text to speech through the OpenAI API. The models are gpt-4o-mini-tts, tts-1 and tts-1-hd, with 13 built-in voices (nine on the two tts-1 models). The guide lists 50+ languages but says the voices are optimized for English, and a Custom Voices feature requires the speaker's consent. OpenAI's usage policies require telling end users the voice is AI-generated.
Pricing: see siteDetails →
Speechify
Reads text aloud with AI voices, and has a separate studio for voiceover and dubbing. The free plan has 10 voices that Speechify itself calls robotic-sounding, at up to 1.5x speed. Premium is $29 per month with 1,000+ voices in 60+ languages.
Free optionDetails →
Voicemod
Real-time voice changer and soundboard for games, streams and calls, with 200+ voices including AI voices trained with professional voice actors. Free-plan limits and prices were not stated on the pricing page when we checked.
Pricing: see siteDetails →
WellSaid Labs
Turns scripts into voiceovers. The free plan gives 3 download minutes per month with no commercial rights. Every paid plan includes commercial usage rights: Starter is $10 per month billed annually (20 download minutes a month), Pro is $33 per month billed annually (180 minutes). WellSaid says its paid voices are English and that customer content is not used to train its models.
Free optionDetails →

Open-weight speech models you run yourself

Models you download and run on your own machine, so there is no per-use bill. The licence is the real filter here: code and weights are often licensed differently.

Chatterbox
Open text-to-speech models from Resemble AI under the MIT license. The repo offers Turbo (350M parameters, English), Nano (110M, English, which it says runs faster than realtime on 8 CPU cores) and Multilingual V3 (500M, 23+ languages). Voice cloning needs a reference clip of about 10 seconds, and every output carries Resemble's Perth watermark. It installs with pip and needs Python 3.11 or newer.
Free optionDetails →
F5-TTS
Open text-to-speech with voice cloning from reference audio. The code is MIT, but the pretrained weights are CC-BY-NC because of the training data, so they are non-commercial. Chinese and English are documented.
Free optionDetails →
Kokoro-82M
An 82 million parameter text-to-speech model under Apache 2.0. Version 1.0 ships 54 voices across 8 languages. It uses a fixed set of voices and does not do voice cloning. Its card says it is built on StyleTTS 2.
Free optionDetails →
Piper
Fast local neural text to speech that runs offline, licensed GPL-3.0. Home Assistant and NVDA use it, and the Open Home Foundation, which now hosts it, says it is looking for maintainers.
Free optionDetails →

How this directory works

Entries are alphabetical inside each group. Position carries no meaning, and a tool being listed is not a recommendation.

We include a tool when it has a public voice or text-to-speech generator and an official page we could open. Each entry shows the date its facts were read from the tool's own site. Where a page did not load or did not state a limit, the entry says so instead of guessing.

A free option here means the tool's own site describes one. It does not tell you how good the free voices are, which is a separate question that our ranked page tries to answer.

Free plans and commercial use

The free plans we read mostly stop at personal use. ElevenLabs, WellSaid Labs, Fish Audio and Murf each say on their own pages that commercial use needs a paid plan or is not allowed on the free one.

Open-weight models are different, and the difference sits in the licence. Kokoro-82M is Apache 2.0, Chatterbox is MIT and Piper is GPL-3.0. F5-TTS publishes its weights under CC-BY-NC, which is non-commercial, even though its code is MIT. Read the licence of the weights, not just the code, before shipping anything.

Choosing a voice tool

Voice tools suit a script you want narrated or a voice you want to reuse. Open-weight models suit volume, privacy or offline use, at the cost of setup time and a machine that can run them.

Missing a tool or spotted a mistake? Tell us.

Frequently asked questions

Is this list ranked from best to worst?: No. Entries are alphabetical inside each group, and the directory makes no quality judgement. Our ranked picks are on the best-of pages, which are written separately.
Has anyone paid to be listed here?: No. Listings are not paid, and the links are plain outbound links marked nofollow. If sponsored placements are ever added, they will be labelled as sponsored and will not change the order of editorial entries.
Can I use audio from a free plan commercially?: Often not. Several free plans in this directory are personal use only, or say commercial use needs a paid plan. The entry for each tool records what its site states, but confirm on the tool's current terms before you publish or sell anything.
Can I clone a voice with these tools?: Several list it. Fish Audio says a clone can be made from a 15-second clip, Cartesia lists instant voice cloning from its Pro plan, ElevenLabs lists it from Starter, and Hume AI's pricing table lists voice cloning on its free plan. Among the open-weight models, Chatterbox and F5-TTS clone from a reference clip, while Kokoro-82M uses a fixed set of voices and does not clone. Only clone a voice when you have the speaker's consent.
Which of these can I run on my own computer?: The models in the open-weight group: Chatterbox, F5-TTS, Kokoro-82M and Piper. Their hardware needs and licences differ, so read each entry. F5-TTS publishes non-commercial weights.
Why are some tools marked 'see site' for free access?: The tool's site did not state its free access in a form we could record, or its pricing page did not load when we checked. We would rather say so than guess.
How can a tool be added or corrected?: Use the contact page. Send the official URL and what the tool does, and we will check the facts on its own site before adding or changing an entry.

Get new guides by email

One email when we publish new guides and model breakdowns. No spam, unsubscribe anytime.