ElevenLabs is the voice generator everyone else gets measured against. It has the most expressive voices and the widest language list, but it isn't cheap. The 2026 alternatives beat it on API price, real-time speed, or cost nothing at all.
Our verdict: Cartesia is the best ElevenLabs alternative for most builders in 2026. Its $5 Pro plan includes a commercial license and instant voice cloning, its Sonic-3.6 model is pitched at real-time voice agents, and its effective API price lands just under ElevenLabs. Murf is the better pick for narrated videos and training content, and its Falcon API is the cheapest per character here. If you just want the best voice quality and don't mind paying for it, ElevenLabs is still the one to beat.
The Short Version
Best ElevenLabs alternative overall: Cartesia. Commercial license and instant cloning from $5 a month, first audio in under 90ms (vendor claim).
Best for voiceovers and e-learning: Murf. A full editing studio from $19 a month billed yearly, plus the cheapest API here at $10 per 1M characters.
Best if you already use OpenAI: gpt-4o-mini-tts. Cheap, steerable with plain-language instructions, but no self-serve voice cloning.
Best open source text to speech: Chatterbox, MIT-licensed with zero-shot cloning in 23 languages. Smallest: Kokoro, an 82M-parameter Apache 2.0 model that runs almost anywhere.
Prices at a glance
Prices checked October 15, 2026, on each company's own pricing page.
| Service | Paid from | API cost per 1M characters | Voice cloning | Languages | Commercial use |
|---|---|---|---|---|---|
| ElevenLabs | $6/mo | $40 (Flash, v4 Turbo) to $80 (v4, v3) | Instant from $6, Professional from $22 | 90+ on v4 | Paid plans |
| Cartesia | $5/mo | About $37 to $39 on Startup and Scale credits | Instant from $5, Pro from $49 | 44 | Paid plans |
| Murf | $19/mo billed yearly | $10 (Falcon), $30 (Gen2) | Enterprise add-on only | 35+ (API) | Paid plans |
| OpenAI TTS | Pay as you go | $15 (tts-1), $30 (tts-1-hd); gpt-4o-mini-tts billed by token | Approved customers only | About 57, tuned for English | Yes, with AI-voice disclosure |
| Chatterbox | Free | Your own hardware | Zero-shot from a short clip | 23 (Multilingual) | Yes, MIT |
| Kokoro | Free | Your own hardware | No | 8 | Yes, Apache 2.0 |
ElevenLabs: the baseline
ElevenLabs is still the most complete voice platform. Its newest model, v4, covers 90+ languages and supports audio tags for fine control over delivery, and v4 Turbo keeps that expressiveness at a claimed ~100ms of latency. Flash, its budget real-time model, claims ~75ms across 32 languages. On the API, Flash, v4 Turbo and v3 Conversational cost $0.04 per 1,000 characters ($40 per 1M), while v4, v3 and Multilingual v2 cost $0.08 ($80 per 1M).
The subscription side is cheap to start. Starter is $6 a month for 30,000 credits with a commercial license and instant voice cloning. Creator, at $22 a month, adds Professional Voice Cloning and 121,000 credits. Pro is $99 for 600,000 credits. The free plan has 10,000 credits but no commercial license.
The catch: heavy use gets expensive fast. Credits are shared across speech, dubbing, music and sound effects, and the top-quality models cost twice as much per character as the fast ones.
Cartesia: the best ElevenLabs alternative
Cartesia is built for speed. Its Sonic-3.6 model claims first audio in under 90ms, it covers 44 languages from one model, and it can clone a voice from about 10 seconds of audio. That makes it a natural fit for voice agents and phone bots.
Pricing undercuts ElevenLabs at the entry level. The $5 Pro plan gives 100,000 credits a month, a commercial license and instant voice cloning. Startup is $49 for 1.25 million credits and adds professional cloning, and Scale is $299 for 8 million. Standard TTS costs about one credit per character, so heavy users pay roughly $37 to $39 per 1M characters, slightly below ElevenLabs' fast models.
The catch: 44 languages is half of what ElevenLabs v4 claims, and Cartesia's tools are aimed at developers. If you want a polished editor for a narrated video, look at Murf instead.
Murf: the best for voiceovers, and the cheapest API
Murf is two products. Murf Studio is a voiceover editor for training videos, presentations and ads, with 200+ voices, pronunciation controls, and Canva and PowerPoint add-ons. Creator costs $19 a month billed yearly ($228) for 24 hours of voice generation a year with commercial rights. Business is $66 a month billed yearly for 96 hours and adds a business license.
The Murf API is the surprise. Falcon, its real-time model, costs $0.01 per 1,000 characters, or $10 per 1M, a quarter of ElevenLabs' cheapest rate. Murf claims time-to-first-audio under 100ms. Its studio-quality Gen2 model costs $30 per 1M. New accounts get $10 of free API credit.
The catch: voice cloning is an Enterprise feature. Studio lists custom voice clones as an Enterprise add-on, and the API offers Professional Voice Cloning only on Enterprise. If you need your own voice, Murf is the wrong tool.
OpenAI TTS: the easy add-on for OpenAI apps
If your app already calls OpenAI, gpt-4o-mini-tts is the simplest upgrade. OpenAI calls it its newest and most reliable speech model, and you can steer accent, tone, emotion and speed with plain-language instructions. It costs $0.60 per 1M text input tokens and $12 per 1M audio output tokens. The older tts-1 costs $15 per 1M characters, and tts-1-hd costs $30.
It has 13 built-in voices and Whisper's language list, though the voices are tuned for English. OpenAI's usage policies require you to tell listeners the voice is AI-generated.
The catch: custom voices are limited to eligible customers through OpenAI's sales team, and you need a consent recording from the speaker. For most people, there's no voice cloning.
Open source text to speech: Chatterbox and Kokoro
Chatterbox, from Resemble AI, is the strongest free option. It's MIT-licensed, and the family now includes Chatterbox Multilingual V3 (500M parameters, 23 languages), Chatterbox-Turbo (350M, English, built for low-latency agents) and Chatterbox-Nano (110M, which Resemble says runs at 3x real time on 8 CPU cores). It clones a voice zero-shot from a short reference clip. Every file carries Resemble's invisible Perth watermark.
Kokoro is the lightweight pick. At 82M parameters with Apache 2.0 weights, it runs on modest hardware, and v1.0 ships 54 voices in 8 languages. It doesn't clone voices, but for a free app voice it's hard to beat.
The catch: you supply the hardware, the setup and the uptime. And check licenses carefully: F5-TTS, another popular cloning model, has MIT code but CC-BY-NC weights, so you can't use its pretrained models commercially.
What about Play.ht?
Play.ht is no longer an option. Meta acquired PlayAI, the company behind it, in July 2025, as reported by Bloomberg and confirmed by Meta, and the play.ht site no longer loads. ElevenLabs and Cartesia are the closest replacements.
How to choose
- Building a voice agent or phone bot: Cartesia, or Murf Falcon if price per character matters most.
- Narrating videos, courses or slides: Murf Studio, or ElevenLabs if you need your own cloned voice.
- Audiobooks and expressive long-form: ElevenLabs v4.
- Already on OpenAI: gpt-4o-mini-tts.
- No budget, or audio must stay on your machine: Chatterbox for cloning, Kokoro for speed.
The verdict
Cartesia is the best ElevenLabs alternative in 2026. It gives you a commercial license and instant voice cloning for $5 a month, a real-time model with a sub-90ms latency claim, and per-character costs slightly below ElevenLabs once you scale.
Pick Murf for voiceover work and the cheapest API, as long as you don't need cloning. Pick Chatterbox if you want free, open source cloning you can run yourself. And if voice quality and language coverage matter more than price, stay with ElevenLabs. If you're wiring voice into an app, our best AI inference API guide covers the models on the other side of the conversation.
FAQ
What is the best alternative to ElevenLabs?
Cartesia is the best ElevenLabs alternative for most people in 2026. Its $5 Pro plan includes a commercial license and instant voice cloning, and its Sonic-3.6 model claims first audio in under 90ms. Murf is the better choice for narrated videos and has the cheapest API at $10 per 1M characters.
Is ElevenLabs better than Murf?
ElevenLabs has more expressive voices, more languages and voice cloning from its $6 Starter plan. Murf has a simpler voiceover editor for training and marketing videos, and its Falcon API costs $10 per 1M characters versus $40 for ElevenLabs' cheapest model. Murf only offers voice cloning on Enterprise.
What is the best text to speech API?
For quality and languages, ElevenLabs is the best text to speech API, at $40 to $80 per 1M characters. Cartesia is the best for low-latency voice agents, Murf Falcon is the cheapest at $10 per 1M characters, and OpenAI's gpt-4o-mini-tts is the easiest choice if you already use OpenAI.
What is the best open source text to speech model?
Chatterbox from Resemble AI is the best open source text to speech model for most uses. It's MIT-licensed, supports 23 languages in its multilingual version and clones voices from a short clip. Kokoro, an 82M-parameter Apache 2.0 model, is the best pick for low-end hardware.
Does Play.ht still exist?
No. Meta acquired PlayAI, the company behind Play.ht, in July 2025, and the play.ht website no longer loads. ElevenLabs and Cartesia are the closest replacements for its voice cloning and API.