I sat down to write an ordinary ElevenLabs alternatives post. Pull the top results, compare the tools, rank them by use case. Standard work.
Then I ran the list of recommended tools through a script that resolves each domain and reports the HTTP status, which took about forty seconds, and two of them came back wrong. Not slow. Not redesigned. One does not exist in DNS. The other returns a status code that means someone stopped paying the hosting bill.
Both are still being recommended in articles published this year. One of the results I had open while researching was suggesting I switch to a product whose parent company is in Chapter 7 liquidation.
So this became a different post. The tools are still ranked by use case further down, and ElevenLabs is still a good product. But the first section is now the part nobody else does.
What I actually checked
On August 3, 2026 I resolved every domain in the recommendation lists and recorded the HTTP status after following redirects. Here is the raw result.
| Tool | Domain | Status | Verdict |
|---|---|---|---|
| ElevenLabs | elevenlabs.io | 200 | Live |
| Murf | murf.ai | 200 | Live |
| Speechify | speechify.com | 200 | Live |
| Resemble AI | resemble.ai | 200 | Live, repositioned |
| WellSaid Labs | wellsaidlabs.com | 200 | Live, owned by Podcastle since 2024 |
| Descript | descript.com | 200 | Live |
| Podcastle | podcastle.ai | 200 | Live |
| Listnr | listnr.ai | 200 | Live |
| Voicemaker | voicemaker.in | 200 | Live |
| Fliki | fliki.ai | 200 | Live |
| Typecast | typecast.ai | 200 | Live |
| Notevibes | notevibes.com | 200 | Live |
| HeyGen | heygen.com | 200 | Live |
| Cartesia | cartesia.ai | 200 | Live |
| Deepgram | deepgram.com | 200 | Live |
| Replica Studios | replicastudios.com | 200 | Resolves, product status unconfirmed |
| Play.ht | play.ht | No response | Dead, domain does not resolve |
| LOVO (Genny) | lovo.ai | 402 | Chapter 7, host returning Payment Required |
Most of the category is fine. Thirteen tools answered normally and are running businesses. This is not a story about AI voice collapsing.
It is a story about two specific entries that should have been removed from every list months ago and were not.
Play.ht is gone, and the domain does not even resolve
Meta acquired PlayAI, the company behind Play.ht, in July 2025. TechCrunch reported it on July 13. The whole team, reported at roughly 35 people, moved into Meta's Superintelligence Labs division to work on Meta AI, AI Characters and wearables, under Johan Schalkwyk, who had previously led voice work at Google.
For customers the sequence went like this. The API went dark in late July 2025, ahead of the announced timeline, which broke integrations for developers who had built on it. New signups closed in August. The service terminated permanently on December 31, 2025. Accounts, voice clones, saved audio and API keys were deleted, and no export tooling was provided.
My check on August 3, 2026 returned NXDOMAIN for play.ht. That is stronger than a dead site. It means the domain has no DNS record at all, so there is nothing to load, no redirect, no farewell page.
I had this in three of my own posts, including the ElevenLabs review that gets more search traffic than anything else I have written. Play.ht was named in that post's excerpt, which means it was sitting in the Google snippet. I fixed all three today. I am including that detail because it would be dishonest to write this piece as though only other people make this mistake. The difference is whether you check.
LOVO is in Chapter 7 liquidation
LOVO, Inc., whose studio product is called Genny, filed for Chapter 7 bankruptcy on May 27, 2026 in the Southern District of New York. The case number is 26-11249. It was reported by TheStreet, MLex, Law360 and Bloomberg Law.
Chapter 7 is liquidation. It is not the version of bankruptcy where a company restructures its debts and continues trading.
The filing arrived while LOVO was defending a copyright lawsuit from a group of voice actors who alleged the company trained its platform on their voices without permission. Bloomberg Law reported that the suit was stayed as a result of the bankruptcy notice, so the actors' claims stopped moving forward when the company filed.
Reporting from July described the site as still selling subscriptions with no bankruptcy notice anywhere on it, and users reporting lockouts and downgraded paid accounts. What I found on August 3 is further along than that: lovo.ai returns HTTP 402 Payment Required. That status code is what a host serves when an account is suspended over billing. The site is not selling anything now because it is not serving anything.
Meanwhile the search result that sent me down this path was recommending LOVO's directable voices as a reason to switch from ElevenLabs.
Two more things the lists get wrong
Resemble AI has repositioned. It appears in nearly every alternatives list as a voice cloning competitor. Reading resemble.ai/pricing today, the plans sold there are for deepfake detection and security: audio and video detection billed per second, identity search per call, watermarking per call, on Flex, Team at $350 a month, Business at $1,000 and custom Enterprise. Voice cloning is not on the pricing page. The generation work is still there, and Resemble builds Chatterbox, but a reader arriving expecting an ElevenLabs-shaped product will not find one at that price structure.
WellSaid Labs is not an independent company. Podcastle acquired it in 2024, and its pricing now runs through Podcastle. It is still listed as a standalone rival in current comparisons. I went looking at this because wellsaidlabs.com and resemble.ai resolve to the same IP address and I suspected Resemble had bought it. That guess was wrong, and I would rather record the wrong guess than quietly present the corrected version as though I knew all along.
One caveat on my own table. Replica Studios returns 200, but it serves from GitHub Pages IP addresses, which is unusual infrastructure for a live commercial SaaS. That is a hint, not a finding. I did not confirm whether the product still works, so I have marked it unconfirmed rather than guessing in either direction.
What happens to your work when a voice tool dies
This is the part I had not thought carefully about until I watched it happen twice, and it changed how I store things.
When Play.ht closed, accounts, saved audio, voice clones and API keys were deleted, and no export tooling was offered. Not degraded or archived behind a paywall. Deleted. LOVO's users, according to reporting from earlier this year, hit a softer version of the same wall: paid accounts downgraded without notice and people locked out of libraries they had paid for, months before the site stopped serving at all.
Three practical consequences follow, and none of them are obvious while a service is working fine.
Your generated audio is only safe if you have it locally. Treating the platform's library as your archive works right up until the day it does not, and the failure mode is total rather than partial. Every render you care about should be downloaded and backed up the day you make it, filed somewhere you control.
Your cloned voice is not portable. When you train a voice on one of these platforms, what you can take away is the audio it produced, not the model that produced it. There is no export format that another vendor will accept. If the platform closes, the clone is gone and you retrain from scratch somewhere else. Which means the thing worth guarding is not the clone, it is the clean source recordings you trained it on. Keep those.
Unused credit is not an asset. It is a promise from a company that might not be there. ElevenLabs, to pick the healthiest example, states that rollover survives only while the subscription stays active and you have not downgraded or cancelled. Buying a large credit buffer to feel safe is the wrong instinct in a category where two vendors just vanished, because the buffer evaporates with the vendor.
The legal risk that never shows up in a pricing table
LOVO did not fail because its product was bad. It filed while defending a lawsuit from voice actors who alleged their voices were used to train the platform without permission, and the bankruptcy stayed that case rather than resolving it. No court decided the underlying question. The company simply stopped existing before anyone had to.
That should shape how you pick a vendor if the voice is going into commercial work, and it is entirely absent from every comparison table I read, mine included until now.
The thing you are buying from a voice vendor is not really audio. It is a licence to use audio commercially, and that licence is only worth as much as the company's right to have trained on whatever it trained on. If that right is contested and the vendor liquidates mid-case, you are left holding output whose provenance nobody ever adjudicated. For a YouTube voiceover that is probably a non-issue. For a national ad campaign, or anything a client will indemnify, it is worth asking about directly before you sign.
So it is worth asking two questions before committing a brand to a voice. Where did the training data come from, and will the vendor put its answer in writing? And is there active litigation, which is public and searchable in a way that a marketing page will never volunteer. I am not a lawyer and this is not legal advice. It is the same instinct as checking whether the domain resolves: cheap to verify, expensive to assume.
Why this keeps happening
Look at who publishes the top results for ElevenLabs alternatives. HeyGen. Murf. Typecast. Notevibes. Brilo. Ringly. eesel. SlideSpeak. Every one of them sells a competing voice product, and the page exists so that their product appears on a list of alternatives.
That incentive does not reward verification. Checking whether the other twelve entries still trade takes time and can only shorten your list. A longer list looks more authoritative. So names get copied from last year's article into this year's, and a product can stay in circulation for eight months after its domain stops resolving.
I do not think anyone is lying. The system just has nobody in it whose job is to check, and readers assume that a post updated for 2026 was looked at in 2026.
The alternatives that are actually alive
Now the ordinary part of the post, restricted to tools that answered on August 3.
For creator voiceover work: Murf or Descript. Murf targets corporate and explainer narration and is polished at it, with fewer natural conversational voices than ElevenLabs and more limited cloning. I could not read Murf's current prices first-hand because murf.ai/pricing renders its plans in JavaScript that my fetch could not see, and third-party pages disagree with each other, quoting everything from $19 to $99. So check the site yourself rather than trusting a number from an article, including this one. Descript is different in shape: voice generation sits inside a full podcast and video editor, and Overdub clones only your own verified voice by design, which is a safety decision rather than a limitation.
For developers who need cheap synthesis: OpenAI TTS. API-first, a handful of preset voices, no cloning, and roughly $15 per million characters, which is dramatically cheaper than ElevenLabs at volume. If your app needs speech output and does not care which specific voice, this is usually the right economic answer.
For real-time voice agents: Cartesia or Deepgram. Both are built around low latency for conversational applications, which is a genuinely different engineering problem from rendering a narration file. ElevenLabs is not the natural pick for a phone agent that has to respond inside a few hundred milliseconds.
For finished video rather than audio files: HeyGen. It outputs presenter-led video, so it is a different product category that happens to compete for the same budget.
For no recurring bill at all: self-hosted Chatterbox. Covered properly below.
The open-source option nobody selling software will recommend
Chatterbox is an open-source TTS model from Resemble AI, MIT licensed, with zero-shot voice cloning and inline emotion tags you place directly in the text, including [laugh], [cough] and [sigh].
I read the repository rather than an article about it. On August 3, 2026: 25,829 GitHub stars, MIT license, most recent commit July 21, 2026, not archived. Several posts describing it quote around 11,000 stars, which is less than half the current figure, so those were written a while ago and never rechecked.
On quality, Resemble published a blind listening study run through Podonos, using identical text and 7 to 20 second reference clips with no prompt engineering or post-processing. The reported result: 65.3 percent preferred Chatterbox, 24.5 percent preferred ElevenLabs, 10.2 percent neutral. An earlier round put it near 63.75 percent.
That is a vendor testing its own model. I am not going to repeat the margin as though it were independent, and neither should anything else you read. What survives the scepticism is the direction: an MIT-licensed model you can run yourself is now close enough to a paid hosted service that the comparison is worth having. In 2023 it was not.
The tradeoff is real and I would rather state it than sell it. You need a GPU. You do the setup and the updates. When it breaks at 2am there is no support queue. What you get is no per-character cost, no credit pool, no plan tiers, and no vendor who can reprice or close on you. Given that this article opened with two companies doing exactly that, the last point is worth more than it used to be.
When ElevenLabs is still the right answer
It usually is, if you are producing content commercially and the voice is the product.
Verified against elevenlabs.io on August 3, 2026: Free is 10,000 credits a month, non-commercial. Starter is $6 for 30,000 credits and is the entry point for commercial use. Creator is $22 for 121,000 and is where Professional Voice Cloning starts. Pro is $99 for 600,000, Scale is $299 for 1.8 million, Business is $990 for 6 million, and Enterprise is custom. Unused credits roll over for two months, so your balance can reach at most three times your monthly quota, provided you do not downgrade or cancel.
Worth noting that several pages currently list Business at $1,320 a month and Scale at 2 million credits. Neither matches what the pricing page says today. The same carelessness that leaves dead products in a list leaves old prices in a table.
Quality on English narration is still the best of the hosted options I have run the same script through, the voice library is enormous, and the company is not going anywhere. My full ElevenLabs review has the pricing math per use case and the credit maths for a weekly upload schedule.
How to check a tool is alive before you pay it
This took me under a minute per tool and I now do it before subscribing to anything.
Load the site and look at the HTTP status rather than the page. A 200 is fine. A 402 or 403 means something is wrong with the account behind it. Nothing at all means the domain is gone.
Find the changelog, blog or release notes and read the most recent date. Products go quiet for months before they close. A newest post from fourteen months ago tells you more than the marketing copy above it.
If there is a public repository, check the last commit date, which is a fact rather than a claim.
Search the company name alongside bankruptcy and acquisition. Court filings and acqui-hires are public, and they routinely predate any notice on the company's own website by months. LOVO's site carried no bankruptcy notice at all while the company was in liquidation, so the website is the last place the news arrives.
If you are budgeting a stack around these tools, my AI tools under $20 a month roundup and the dated record of every 2026 AI pricing change both apply the same check-it-yourself approach to prices, and the YouTube automation playbook shows where voice sits in a working production stack.
Sources, checked August 3, 2026:
- Domain and HTTP status for all 18 tools: resolved directly, results in the table above
- TechCrunch: Meta acquires voice startup Play AI
- TheStreet: AI voice cloning firm collapses in Chapter 7 bankruptcy
- MLex: Lovo files for Chapter 7 amid AI voice-cloning lawsuit
- Bloomberg Law: Voice actors' suit against Lovo paused amid bankruptcy
- LOVO, Inc. Chapter 7, case 26-11249
- Chatterbox repository and Resemble's Chatterbox page
- ElevenLabs pricing and Resemble AI pricing

