Voice AI

ElevenLabs Alternatives: 8 Voice AI Tools Compared (2026)

Compare 8 ElevenLabs alternatives for text to speech, voice agents and speech to text, with verified November 2026 pricing and MCP support.

Sukanya Kakoty·Published on 20 Nov 2026 · 9 min read
Share
ElevenLabs alternatives 2026: 8 voice AI tools including Murf, Cartesia, Async, Vapi, Retell AI, Bland AI, Deepgram and AssemblyAI

Last verified: November 18, 2026 (every price below was checked on the vendor's own site or docs on this date)

ElevenLabs does three different jobs: text to speech (TTS), voice agents that hold phone or web conversations, and speech to text (STT). Most people searching for "ElevenLabs alternatives" only need one of those jobs, and a TTS tool is the wrong swap if you're actually building a phone agent. This list splits the 8 strongest ElevenLabs competitors into those three buckets so you compare like with like.

Compare ElevenLabs alternatives on Maarket →

Quick Answer

As of November 2026, the best ElevenLabs alternative depends on which ElevenLabs job you're replacing. For text to speech, pick Murf (Falcon API at $0.01 per minute), Cartesia (paid plans from $5/month) or Async, formerly Podcastle (from $19.99/month). For voice agents, pick Vapi ($0.05/min platform fee), Retell AI ($0.07-$0.31/min) or Bland AI ($0.12-$0.14/min). For speech to text, pick Deepgram (Nova-3 from $0.0048/min) or AssemblyAI (from $0.15 per audio hour). Skip Play.ht: it shut down on December 31, 2025.

ElevenLabs alternatives compared

ElevenLabs baseline for reference: Free (10,000 credits/month), Starter $6/month, Creator $22/month, Pro $99/month, Scale $299/month, Business $990/month. Annual billing costs 10x the monthly price. ElevenLabs has an official MCP server. Sources: ElevenLabs pricing, ElevenLabs API pricing, ElevenLabs MCP server.

Swipe sideways to see all columns →

#
ToolBucketStarting price (Nov 2026)Best forAgent-ready / MCP
1
Murf AIText to speechFalcon API $0.01/min; Studio free, paid from $29/moVoiceovers plus cheap real-time TTSOfficial MCP server (murf-mcp)
2
CartesiaText to speech (also STT, agents)Free; Pro $5/mo; Startup $49/moLow-latency TTS for voice agentsOfficial hosted MCP (mcp.cartesia.ai)
3
Async (formerly Podcastle)Text to speech for creatorsFree; Essentials $19.99/mo; Voice API from $0.50/hrPodcasters and video creatorsVoice API; no official MCP found
4
VapiVoice agents$0.05/min platform fee + providers at costDevelopers who want to pick every providerOfficial MCP (mcp.vapi.ai)
5
Retell AIVoice agents$0.07-$0.31/min, $10 free creditsSupport and sales calls with itemized pricingOfficial MCP (mcp.retellai.com); agents can call MCP tools
6
Bland AIVoice agentsStart: $0 platform fee, $0.14/min; Build $299/mo, $0.12/minHigh-volume phone callingOfficial MCP, 39 tools (api.bland.ai/v1/mcp)
7
DeepgramSpeech to text (also TTS, agents)Nova-3 $0.0048/min; $200 free creditReal-time transcription inside agentsOfficial MCP via Deepgram CLI + docs MCP
8
AssemblyAISpeech to text (new TTS)From $0.15/audio hr; $50 free creditsTranscription-heavy productsOfficial MCP (mcp.assemblyai.com)

Every tool on this list has a public API. "Official MCP" means the vendor publishes and documents its own Model Context Protocol server, so agents like Claude, Cursor or ChatGPT can operate the account directly.

Why this list matters

Many 2026 roundups of ElevenLabs alternatives are already out of date. Play.ht still shows up as a pick even though Meta acqui-hired the PlayAI team in July 2025 and the service shut down on December 31, 2025. Podcastle rebranded as Async on January 28, 2026. Bland's pricing page now lists three plans (Start, Build, Enterprise), not four, and Cartesia's Scale plan is $299/month. We built this list from scratch, checked every price and MCP claim on the vendor's own site on November 18, 2026, and cut anything we couldn't verify.

Sources for the Play.ht shutdown and Podcastle rebrand: Murf's Play.ht transition guide, Podcastle rebrands as Async.

The 8 best ElevenLabs alternatives, by bucket

Text to speech alternatives

Text to speech tools turn written text into audio files or live audio streams. They replace ElevenLabs' core voice generator.

1. Murf AI

What it is: An AI voiceover studio for narration, e-learning and marketing videos, plus Falcon, a real-time TTS API built for voice agents.

Pricing: The Falcon API costs $0.01 per minute. Murf Studio has a free plan, with paid plans from $29/month on monthly billing (Business $99/month) and custom Enterprise pricing.

Standout: Murf claims 55 ms model latency and about 130 ms end-to-end latency for Falcon, with 150+ voices across 35 languages. At $0.01/min, Falcon costs about 14x less per minute of speech than ElevenLabs Creator on Multilingual v2 (about $0.14/min on the same 775-characters-per-minute basis). Murf also publishes an official MCP server.

Best for: Teams that need both a polished narration studio and a low-cost TTS API for agents.

Sources: Murf API pricing, Murf pricing, Murf MCP server docs

2. Cartesia

What it is: A real-time voice AI platform whose Sonic TTS models are built for low-latency conversation. It also sells Ink speech to text and managed voice agents.

Pricing: Free (20,000 credits/month), Pro $5/month (100,000 credits, commercial license), Startup $49/month (1.25M credits), Scale $299/month (8M credits), Enterprise custom. TTS costs 1 credit per character, about 750-800 credits per minute of audio. Managed agent calls cost $0.03/min plus $0.014/min for a Cartesia phone number, with LLM usage billed separately.

Standout: Cartesia's $5/month Pro plan is the lowest-priced paid subscription in this list that includes a commercial license and an official hosted MCP server at mcp.cartesia.ai/mcp.

Best for: Developers building voice agents who want a fast TTS layer they can plug into Vapi, Retell or their own stack.

Sources: Cartesia pricing, Cartesia MCP docs

3. Async (formerly Podcastle)

What it is: Podcastle rebranded as Async on January 28, 2026. It's an all-in-one creator platform for recording, editing, AI voices and voice cloning, plus a developer Voice API.

Pricing: Creator plans are Free, Essentials $19.99/month ($143.90/year), Pro $39.99/month ($287.90/year), Teams $84.99/month and custom Business. The Voice API charges $0.50 per hour of generated audio (Flash) or $1.00 per hour (Pro), with instant voice cloning included.

Standout: At $0.50 per generated hour (about $0.008 per minute), Async's Flash API has the lowest TTS rate we verified, and TTS sits inside a full recording and editing workflow. Async has a REST and WebSocket API, but we found no official MCP server.

Best for: Podcasters, YouTubers and small content teams who want TTS inside their editor rather than as a separate tool.

Sources: Podcastle rebrands as Async, Async Creator plans, Async Voice API pricing

Voice agent alternatives

Voice agent platforms run full conversations: they listen, call an LLM, speak back and connect to phone lines. They replace ElevenLabs' agents product, and most of them can still use ElevenLabs voices.

4. Vapi

What it is: A developer platform for building phone and web voice agents. You choose the transcriber, LLM, voice and telephony provider, and Vapi orchestrates the call.

Pricing: A flat $0.05 per minute platform fee, prorated to the second, with transcriber, model, voice and telephony costs passed through at provider cost. Optional packages: Core $29/month (support, 10 concurrent calls) and Pro at 10% of the hosting fee with a $999/month minimum. HIPAA is a $2,000/month add-on.

Standout: Vapi has the most provider choice in this list, including ElevenLabs voices and Deepgram transcription, so you can keep ElevenLabs as the voice and swap everything around it. It also has an official MCP server at mcp.vapi.ai/mcp.

Best for: Engineering teams who want full control over every layer of their voice agent stack.

Sources: Vapi pricing, Vapi pricing docs, Vapi MCP server

5. Retell AI

What it is: A voice agent platform for inbound and outbound calls, with itemized per-component pricing and built-in testing and analytics.

Pricing: Pay-as-you-go voice agents cost $0.07-$0.31 per minute. That breaks down into Retell Voice Infra at $0.055/min, platform voices from $0.015/min (ElevenLabs voices $0.04-$0.10/min) and the LLM you pick. New accounts get $10 in free credits and 20 free concurrent calls, with extra concurrency at $8 per call per month.

Standout: Retell publishes a rate for every component, so you can model cost per call before you build. Its official MCP server (mcp.retellai.com) lets coding agents build and test Retell agents, and Retell agents can call external MCP servers as tools.

Best for: Support and sales teams who want a production phone agent quickly without stitching providers together.

Sources: Retell AI pricing, Retell MCP server docs

6. Bland AI

What it is: An enterprise phone agent platform for running large volumes of AI calls, inbound or outbound.

Pricing: Start has no platform fee and charges $0.14/min of talk time. Build costs $299/month at $0.12/min. Enterprise is custom. Telephony is billed separately at pass-through cost, and call transfers add $0.04-$0.05/min on self-serve plans.

Standout: Enterprise offers unlimited calls and on-prem or VPC deployment. Bland's official MCP server exposes 39 tools for building agents, placing calls and running evals.

Best for: Companies with high, predictable call volume and compliance requirements.

Sources: Bland AI pricing, Bland MCP server docs

Speech to text alternatives

Speech to text APIs transcribe audio into text, live or after recording. They replace ElevenLabs' transcription product.

7. Deepgram

What it is: A voice AI API platform best known for speech to text (Nova-3 and the conversational Flux model). It also sells Aura-2 TTS and a Voice Agent API.

Pricing: Pay-as-you-go Nova-3 monolingual costs $0.0048/min (about $0.29 per hour), Flux English $0.0065/min, Aura-2 TTS $0.030 per 1,000 characters, and the Voice Agent API $0.075/min (Standard). New accounts get $200 in free credit with no card required. Growth needs $4,000/year in prepaid credits. Deepgram labels some streaming rates as promotional, so check before you budget.

Standout: Flux is built for turn-taking in live conversations, and Deepgram's $200 starting credit is the largest free credit in this list. Deepgram ships an official MCP server through its CLI (dg mcp).

Best for: Developers who need fast, accurate real-time transcription inside a voice agent.

Sources: Deepgram pricing, Deepgram CLI MCP server

8. AssemblyAI

What it is: A speech AI API for transcription and audio intelligence (speaker labels, summaries, entity detection). It launched AssemblyAI-TTS-1 on October 27, 2026.

Pricing: New accounts get $50 in free credits. Universal-2 pre-recorded and Universal-Streaming each cost $0.15 per audio hour, and Universal-3.6 Pro Realtime costs $0.45 per hour. TTS-1 costs $15 per million characters. AssemblyAI is replacing its Universal models with a new model family named AssemblyAI-STT-3.7 and is retiring its Voice Agent API, so confirm the current model and rate before you build.

Standout: $0.15 per hour is the lowest STT rate we verified, and the official MCP server at mcp.assemblyai.com/mcp lets agents transcribe and analyze audio on your account.

Best for: Products built on transcription at volume, such as meeting notes, call analytics and media search.

Sources: AssemblyAI pricing, AssemblyAI pricing reference

How to choose an ElevenLabs alternative

Start with the job, then the price. Ask one question first: what does the audio do?

  1. It's read once and published (voiceovers, audiobooks, ads, course narration): you need text to speech. Shortlist Murf, Cartesia or Async.
  2. It talks back to a person in real time (support lines, booking calls, sales calls): you need a voice agent platform. Shortlist Vapi, Retell AI or Bland AI. These platforms can still use ElevenLabs voices.
  3. It listens and writes things down (transcripts, captions, call analytics): you need speech to text. Shortlist Deepgram or AssemblyAI.

Then compare effective cost per minute, not plan price. Here's what one minute of generated speech costs on each TTS option we verified:

TTS optionEffective cost per minute of speech
ElevenLabs Creator, Multilingual v2 ($22 for 121,000 credits)~$0.14
ElevenLabs Creator, low-latency models (half the credits per character)~$0.07
Cartesia Startup ($49 for 1.25M credits)~$0.03
Deepgram Aura-2 ($0.030 per 1,000 characters)~$0.023
AssemblyAI TTS-1 ($15 per 1M characters)~$0.012
Murf Falcon API (published rate)$0.01
Async Flash API ($0.50 per hour)~$0.008

Our arithmetic uses one basis for every tool: about 775 characters per minute of speech, the midpoint of Cartesia's published 750-800 credits per minute. ElevenLabs Multilingual v2 uses 1 credit per character, and its low-latency models give twice the minutes per credit, according to ElevenLabs' API pricing page. ElevenLabs' own estimate (about 10 minutes per 10,000 credits) assumes a longer script per minute, so it looks slightly more expensive than our figure.

Last, check whether the tool is agent-ready. If you plan to let an AI agent build, test or run your voice workflows, prefer vendors with an official MCP server. Seven of the 8 tools here have one, and so does ElevenLabs. For more options across the category, browse Voice AI tools on Maarket.

How we built this list

We grouped ElevenLabs' products into three jobs (text to speech, voice agents and speech to text), then searched for the strongest current alternative in each, including shutdowns and rebrands. We checked each tool's pricing, free tier and MCP support on its own website or documentation on November 18, 2026. Where a vendor's pages disagreed (Vapi lists different free-credit amounts on its pricing page and its docs), we left the figure out. We removed Play.ht because the service shut down on December 31, 2025, and swapped in AssemblyAI. Effective per-minute costs are our own arithmetic from published rates. Prices change often, so confirm on the vendor's site before you buy.

Last reviewed: November 18, 2026

Frequently asked questions

What is the best ElevenLabs alternative in 2026?

It depends on which ElevenLabs product you're replacing. For text to speech, Murf (Falcon API at $0.01/min) and Cartesia (from $5/month) are the strongest picks. For voice agents, choose Vapi, Retell AI or Bland AI. For speech to text, choose Deepgram or AssemblyAI. All 8 have public APIs, and 7 have official MCP servers.

Is there a free ElevenLabs alternative?

Yes, most have free tiers. Cartesia's Free plan includes 20,000 credits a month, Murf Studio and Async have free plans, Deepgram gives $200 in free credit, AssemblyAI gives $50, and Retell gives $10. Free tiers usually don't include a commercial license: Cartesia's commercial license starts on its $5/month Pro plan.

What is the cheapest ElevenLabs alternative for text to speech?

By published API rates, Async's Flash Voice API is the cheapest we verified at $0.50 per hour of generated audio (about $0.008 per minute). Murf Falcon is next at $0.01 per minute. ElevenLabs Creator works out to about $0.14 per minute on Multilingual v2 and about $0.07 on its low-latency models, so Async and Murf are roughly 14-17x cheaper than Multilingual v2 and 7-8x cheaper than ElevenLabs' low-latency models.

How much does ElevenLabs cost in 2026?

As of November 2026, ElevenLabs has a Free plan with 10,000 credits a month. Paid plans are Starter at $6/month, Creator at $22/month, Pro at $99/month, Scale at $299/month and Business at $990/month, plus custom Enterprise pricing. Annual billing costs 10 times the monthly price, which works out to two months free.

Can I keep ElevenLabs voices and switch platforms?

Yes, if you're switching voice agent platforms. Vapi lets you pick ElevenLabs as the voice provider and passes the cost through with no markup. Retell AI offers ElevenLabs voices at $0.04-$0.10 per minute, depending on the model. You keep the voice and change the orchestration, telephony and LLM around it.

Which ElevenLabs alternative has the lowest latency?

Murf and Cartesia publish the lowest vendor latency figures in this list. Murf claims 55 ms model latency and about 130 ms end-to-end for Falcon. Cartesia claims under 90 ms time to first audio for Sonic, and independent benchmarks have measured roughly 130-190 ms once network time is included. Vendor numbers measure different things, so test from your own region before you choose.

Is Play.ht still available?

No. Meta acqui-hired the PlayAI team (Play.ht's parent company) in July 2025, and the Play.ht service shut down on December 31, 2025. Former Play.ht users looking for a similar TTS API should compare Murf, Cartesia and Async.

Building a voice AI product?

Maarket ranks voice AI tools by Karma Points: real usage signals like reviews, "I use this" clicks and recommendations, plus agent-readiness and MCP support. Rankings update continuously instead of resetting after a one-day launch.

Browse the Voice AI category on Maarket →

List your voice AI product on Maarket →

Written by

Share this article

Voice AI products on Maarket

  • ElevenLabs — ElevenLabs is an AI voice generator that makes lifelike speech and audio for creators and teams.
  • Suno — Suno is an AI music generator that turns a text prompt into full songs with vocals for creators.
  • Otter.ai — Otter.ai is an AI meeting assistant that transcribes and summarizes meetings for teams.
  • Yap — Hold a key, speak, and Yap types into any app. Speech never
  • Voiskey — AI voice typing that sounds right in every app
  • MosMos — Voice writing that works before, during, and after meetings
  • Gemini 3.8 & 3.8 Live Extended Thinking — Our most advanced Gemini Audio models yet
  • NovaSynth by Noveum — Test your voice agent on the callers you can't stage.

Keep reading

The best tools, in your inbox

One email a week — new rankings, fresh comparisons, and what the community is switching to.