If you are choosing an AI voice tool for narration, training videos, ads or an app, ElevenLabs, Murf and Play.ht are three names that come up often. This guide is a decision framework plus researched facts about each product, not a ranking. One fact changes the premise of the title, so we start with it: Play.ht is no longer a live option for new customers.
The aim is to help you decide what to compare, in what order, and what evidence to collect before you pay. We name the differences the vendors document, flag the gaps we could not close, and leave the final call to your own listening tests.
How we researched this: we did not run hands-on tests for this article. It is based on the vendor’s own documentation and public sources, checked on 11 November 2026. Features and prices change, so confirm details on the vendor’s site before you buy.
Key takeaways
- Play.ht (also called PlayHT or PlayAI) was reported as acquired by Meta in July 2025, and a September 2026 third-party article says it shut down on 31 December 2025. Treat it as a historical name, not a shortlist candidate.
- ElevenLabs documents the broadest platform of the three: several text to speech models, speech to text, dubbing, music, agents and an API, with 90+ languages on its newest models.
- Murf presents itself as a studio for voiceovers, with a separate pay as you go API, a dubbing product and a stated 200+ voices and 20+ languages for dubbing and translation.
- Voice cloning, commercial rights and data handling vary by plan on both live products, so these are the criteria to verify before you commit.
- We could not read either vendor’s live pricing table, so this guide describes pricing models only and quotes no plan prices. Run a short listening test with your own script.
Why compare these, and what changed
AI voice tools turn text into spoken audio. They are used for explainer videos, e-learning, podcasts, audiobooks, phone agents and product demos. The three names in the title each sit in a slightly different place: ElevenLabs as a wide audio platform, Murf as a voiceover studio for business content, and Play.ht as a former text to speech service. If you are new to the topic, our beginner’s guide to AI voice and text to speech explains the basics.
The Play.ht premise has changed
Biometric Update reported on 15 July 2025, citing Bloomberg and an internal Meta memo, that Meta had acquired PlayAI and that the whole team would join Meta. Financial terms were not disclosed, and that article did not say what would happen to the product.
Later sources are more direct but less authoritative. Murf published a transition guide on 11 February 2026 describing Meta’s acquisition of PlayHT and saying the platform was preparing to shut down by 31 December 2025, although the post date is after that date, so its wording is inconsistent. An AI Comparison article dated 7 September 2026 says PlayHT shut down completely on 31 December 2025, that its API stopped accepting requests around 26 July 2025, and that new sign ups ended in August 2025. We could not open an official shutdown notice from the vendor, so we present this as reported, not confirmed by Play.ht itself.
In practice, this means the useful comparison today is between ElevenLabs and Murf, with Play.ht relevant only if you have old audio, old API integrations or old voices to migrate. Murf says it offers recent PlayHT subscribers a free six month subscription that matches or upgrades their previous plan, on proof of subscription, plus migration support and an API migration guide. That is Murf’s own offer, so check its current terms.
Quick overview table
The table below uses only statements we read on vendor documentation or vendor blogs, with the source noted. Blank or unclear items are marked as not confirmed.
| Item | ElevenLabs | Murf | Play.ht |
|---|---|---|---|
| Current status | Live, active product updates (Oct 2026 blog entries) | Live, active blog and API pages | Reported shut down 31 Dec 2025 (third party) |
| Product shape | ElevenCreative web app, ElevenAgents, ElevenAPI, Reception.ai | Murf Studio, Dubbing, Voice Changer, API (Falcon, Gen 2), voice agents | Text to speech platform, now reported closed |
| Languages stated | 90+ on Eleven v4 and Scribe v3; 70+ on v3; 32 on Flash v2.5; 29 on Multilingual v2 | 20+ for dubbing and translation (Feb 2026 blog) | Not confirmed |
| Voice library stated | Over 10,000 community voices (docs); 17,500+ voices (blog) | 200+ AI voices (Feb 2026 blog) | Not confirmed |
| Voice cloning | Instant and Professional cloning, plus voice design from text | Clone your voice (blog); one third party says Enterprise only | Not confirmed |
| Speech to text | Scribe models | Speech to text and audio to text tools listed | Not confirmed |
Note the mismatch in ElevenLabs numbers: its documentation says the Voice Library holds over 10,000 community shared voices, while its blog cites 17,500+ voices. The two may count different things, and we could not reconcile them from the pages. Murf’s blog likewise says 200+ voices while a third party says 120+ standard voices, rising past 200 with the Gen 2 model. Always count voices in the live product for the language and accent you need.
Criteria that matter
Before comparing features, decide which of these actually decide your project. A short list keeps the comparison fair, which is the approach in our guide to comparing AI tools side by side fairly.
- Voice quality for your content type. Narration, dialogue and ads stress voices differently. Only your own listening test settles this.
- Languages and accents you need. Counts on a marketing page do not tell you whether your specific language sounds natural.
- Cloning and consent. Do you need a custom voice, and who is allowed to supply the sample?
- Licence and commercial use. Can you use the audio in paid ads or sell it inside a product on your plan?
- Workflow fit. Do you need a timeline editor for video, or an API for an app?
- Data handling. What does the vendor keep, and can you limit it?
- Cost model. Time based, credit based or per character, and what a revision costs.
Voices, models and languages
ElevenLabs
ElevenLabs documentation lists several text to speech models. Eleven v4 is described as the most expressive, with 90+ languages and a 10,000 character limit per request. Eleven v4 Turbo is the real time option, with about 100 ms median latency. Eleven v3 covers 70+ languages with a 5,000 character limit. Eleven Multilingual v2 is described as the most stable for long form, with 29 languages. Eleven Flash v2.5 is described as fast and lower cost, with 32 languages and a 40,000 character limit.
The practical lesson is that “ElevenLabs” is not one voice engine. Language count, length limit and speed depend on the model you pick, so when you test, record which model produced each sample. One of its own pages also states “29+ languages” for the Text to Speech API while the main description cites 90+, and we could not reconcile this.
Murf
Murf describes Murf Studio as its voiceover product and Murf Gen 2 as a speech model with precise voice controls. Its February 2026 blog lists 200+ AI voices, plus 20+ languages for dubbing and translation. A third party article from September 2026 describes script editing, pronunciation controls, background music and a timeline that syncs narration to video or slides. We treat those workflow details as reported, since we could not open Murf’s product pages in full.
| Question to answer | Where to look | What we found |
|---|---|---|
| Which model does my sample use? | ElevenLabs model list in docs | Language support and limits differ by model |
| How many voices fit my language? | Live voice browser in each product | Headline counts differ and are not like for like |
| Can I control pronunciation? | Studio editor or API settings | Murf is reported to offer pronunciation controls; check ElevenLabs equivalents in docs |
| Is long form stable? | Your own 5 to 10 minute script | ElevenLabs describes Multilingual v2 as most stable for long form |
Voice cloning and consent
ElevenLabs documents two cloning approaches. Instant Voice Cloning creates a voice from short samples, under two minutes, and is available on most plans. Professional Voice Cloning uses longer training audio for higher fidelity and requires a Creator plan or above. Professional clones go through a verification step the docs call voice captcha technology, which confirms they come from your own samples. It also lets you design a voice from a text description. In the Voice Library, Professional clones can be shared publicly while Instant clones cannot, and free tier users cannot use Voice Library voices through the API.
Murf’s blog says “Clone your voice to create accurate voiceovers for any project.” A September 2026 third party review says cloning is available only on the Enterprise plan, and a separate October 2026 article lists cloning only under Enterprise while leaving its own FAQ answer blank. We could not confirm the current plan rules from Murf itself, so ask Murf directly if cloning matters.
Cloning raises consent and legal questions whichever tool you use. Only clone voices you have clear written permission to use, and read our guide to voice cloning consent and legal basics before you start.
Workflow, dubbing and API
ElevenLabs lists ElevenCreative as a no code web app for voiceovers, music, dubs and studio projects, ElevenAPI as a REST interface with official Python and TypeScript SDKs, and ElevenAgents for voice agents. Its docs list Scribe v3 for batch speech to text and Scribe v3 Live for real time. It also notes that music_v1 will be removed on 31 January 2027 and eleven_text_to_sound_v2 on 1 March 2027, a reminder to check deprecation notices if you build on any API.
Murf lists Murf Dubbing for translating video and audio, a Voice Changer that turns recordings into voiceovers, an Accent Generator, and integrations with Canva, Google Slides, PowerPoint, Adobe Captivate and Windows. For developers it lists Murf Falcon, described as a text to speech API built for voice agents and low latency, plus Gen 2, Voice Changer, Dubbing Automation and Translation APIs.
| Workflow need | ElevenLabs documents | Murf documents |
|---|---|---|
| Browser voiceover editor | ElevenCreative | Murf Studio |
| Dubbing | Dubbing and studio projects (in the product list) | Murf Dubbing, 20+ languages |
| Developer API | ElevenAPI with Python and TypeScript SDKs | Falcon, Gen 2 and other APIs |
| Presentation tools | Not stated on pages we read | Canva, Google Slides, PowerPoint, Adobe Captivate |
| Voice agents | ElevenAgents | AI voice agent products |
Pricing models
We could not read the live pricing tables on either vendor’s pricing page: Murf’s page returned only a description to our reader, and ElevenLabs’ pricing page was not accessible to us. So we do not quote plan prices. Describing the models is still useful, and our guide on how to read an AI tool’s pricing page shows what to check.
- ElevenLabs: its blog says commercial rights vary by subscription tier, and its docs tie Professional Voice Cloning to a Creator plan or above. It also advertises limited time offers, so check what you are actually buying and for how long.
- Murf: a third party article dated 5 October 2026, which says it took its Studio figures from Murf’s pricing page, describes Free, Creator, Business and Enterprise tiers measured in voice generation time, with commercial rights on paid tiers and cloning only on Enterprise. It also says the API is billed separately by character count, so a Studio plan does not cover API use. Confirm all of this on Murf’s page.
- Play.ht: no longer relevant for new buyers if the reported shutdown is accurate.
When you compare, convert every plan into the same unit, for example cost for ten minutes of finished audio including two revisions. Time allowances, credits and characters are not interchangeable.
Privacy and data
ElevenLabs offers Zero Retention Mode as an enterprise option. It restricts logging of inputs, outputs and the account email for eligible API requests, keeps request data only for processing, and is switched on per request with enable_logging=false or per agent. It is for select enterprise customers, mainly in healthcare and banking, and access can be restricted. It covers Text to Speech, Text to Dialogue, Voice Changer, Speech to Text and Sound Effects endpoints. It does not cover web UI traffic, dubbing, ElevenCreative Studio, voice cloning or music fine tunes, and it makes troubleshooting harder.
Murf states it is ISO/IEC 42001, ISO 27001, SOC 2 Type II, CCPA and GDPR compliant, with customer data stored in AWS US-East-2 (Ohio), AES-256 encryption at rest and TLS 1.2 or higher in transit. After an agreement ends there is a grace period of up to 90 days to retrieve generated audio, and deletion requests are completed without undue delay, with complex cases up to 90 days. Its security page does not say whether customer content is used to train models, so check its privacy policy and terms.
For any vendor, avoid pasting confidential scripts into a free tier until you have read the data terms. Our guide to protecting company data when using AI tools gives a checklist.
Who each suits
| If you need | Documented fit |
|---|---|
| Many models, many languages, speech to text and an API in one account | ElevenLabs documents the widest range of products and models |
| Voiceovers for slides, training or marketing video with a studio editor | Murf documents a studio, presentation integrations and dubbing |
| Real time voice agents | Both document agent products or low latency models; test latency yourself |
| Strict enterprise data controls | ElevenLabs Zero Retention Mode (enterprise, with limits); Murf lists SOC 2 Type II and ISO certifications |
| To keep using an old Play.ht integration | Plan a migration; the product is reported closed |
This is a “best for X if Y” reading of documented features, not a score. Someone who needs a simple voiceover for a slide deck may not need an agent platform, and a developer building a voice app may find a studio editor irrelevant.
How to run your own test
Use the method in our guide on testing an AI tool before you pay. A simple version for voice tools:
- Pick one script of about 300 words that includes numbers, product names, a question and a list.
- Generate it in each tool with the closest voice style and record the model or voice name used.
- Listen blind: ask two colleagues to rank clips without knowing which tool made them.
- Count fixes needed: mispronounced names, odd pauses, wrong stress. Note how long each fix took.
- Repeat in a second language if you publish in more than one.
- Check the licence wording for your use case on the plan you would actually buy.
Decision checklist
- Is the product live and are you buying a current plan, not a legacy one?
- Does your plan allow commercial use for your exact use case?
- Do you need cloning, and does your plan include it?
- Have you checked the languages you need by listening, not by counting?
- Do you know whether API use is billed separately from the studio plan?
- Have you read the data and retention terms, including training use?
- Can you export your audio and move away if prices or terms change?
Frequently asked questions
Is Play.ht still available?
Reports say no. Biometric Update reported Meta’s acquisition of PlayAI on 15 July 2025, and a third party article dated 7 September 2026 says the service shut down on 31 December 2025. We could not open an official notice, so confirm with the vendor if you rely on it.
Which is better, ElevenLabs or Murf?
We do not rank them. ElevenLabs documents a wider platform with many models and languages, while Murf documents a studio focused on voiceovers with presentation integrations. Which fits depends on your workflow and your listening test.
Can I use the audio commercially?
It depends on your plan. ElevenLabs says commercial rights vary by subscription tier, and a third party says Murf’s free plan has no commercial rights. Read the licence for your exact plan.
Do both support voice cloning?
ElevenLabs documents Instant and Professional cloning. Murf says you can clone your voice, but a third party reports it only on Enterprise. Confirm directly with Murf.
Is my text used to train the vendor’s models?
We could not confirm this for Murf from its security page. ElevenLabs documents Zero Retention Mode for enterprise API use, with the limits above. Read each vendor’s privacy policy for your plan.
What should I do if I used Play.ht before?
Export any audio you still have, list the voices and API calls you depended on, and test replacements with the same scripts. Murf says it offers migration help and a free six month subscription for recent subscribers, subject to its terms.
Conclusion
The honest comparison today is ElevenLabs versus Murf, with Play.ht a reported closed product. Choose by documented fit: platform breadth and models on one side, a voiceover studio with presentation tools on the other. Then verify languages, cloning rules, commercial rights and data terms with a short test. For the method, see how to compare AI tools side by side fairly, and for the money side see how to read an AI tool’s pricing page.
Sources
- ElevenLabs documentation: overview, accessed 11 November 2026
- ElevenLabs documentation: voices and cloning, accessed 11 November 2026
- ElevenLabs documentation: Zero Retention Mode, accessed 11 November 2026
- ElevenLabs blog: PlayHT alternatives, accessed 11 November 2026
- Murf blog: Play AI transition guide (11 February 2026), accessed 11 November 2026
- Murf: Security and trust, accessed 11 November 2026
- Biometric Update: Meta acquires PlayAI (15 July 2025), accessed 11 November 2026
- AI Comparison: Murf AI vs PlayHT (7 September 2026), accessed 11 November 2026
- Smallest.ai: Murf AI pricing (5 October 2026), accessed 11 November 2026
