PlayHT

What it actually does

Nothing, as far as we can tell. That is the only thing on this page that matters.

Meta acquired PlayAI in July 2025, confirmed on 11 to 13 July by Bloomberg and TechCrunch, with the entire team moving to Meta and financial terms undisclosed. TechCrunch's Anthony Ha wrote on 13 July 2025 that "Meta has acquired Play AI, a startup that uses AI to generate human-sounding voices." Third-party trackers then reported that play.ht was permanently shut down on 31 December 2025, with no new signups and no generation. That shutdown date comes from a third-party source (Notevibes) rather than a first-party announcement, so treat it as strong but not confirmed by the vendor. The practical position is the same either way: do not buy this, and if a page is still taking card details, do not give it any.

For the record, what it used to be was a text-to-speech platform with a very large voice library, real-time generation, voice cloning and a developer API, covering 142 languages. Developers liked the low-latency API and the accent coverage, and reviewers consistently praised voice realism and range.

What it costs

The advertised pricing was 12,500 free characters with no commercial rights, Creator at $39 a month, Pro at $99 a month, and Enterprise on request. Those numbers are historical. The only cost that matters now is the one you should not pay.

The catch, if you are an existing subscriber, is the obvious one. A discontinued product means your voices, your cloned models and your generation pipeline stop being available, and there is no roadmap to appeal to. If you have an API key in production, treat that integration as already dead and move it.

Where it falls down

The service was already in trouble before the acquisition landed, which is worth knowing because it tells you the ending was not purely about Meta's interest in the team.

Support was the defining complaint. G2's own auto-summary flagged a lack of responsive customer support, and the most recent reviews on the profile were 0 out of 5 ratings centered on exactly that. One G2 reviewer in July 2025 titled their zero-rated review around good product, nightmare customer support, which summarizes the pattern neatly. Users also reported platform unreliability, with constant connection errors and voiceovers that simply failed to generate, and said the pricing was high relative to output and to what competitors offered.

The review scores read better than the product's ending suggests, and the gap is instructive. G2 showed 4.2 out of 5 from 89 reviews, with 10 percent of those at one star. AppSumo showed 4.18 from 114 reviews, mostly lifetime-deal buyers. The successor domain play.ai sits at 2.2 from nine reviews on Trustpilot, which is far too small a sample to mean much on its own but points in the same direction as the late G2 reviews. Averages built over years of a working product do not update quickly when the product stops working.

That is how a tool with a 4.2 on G2 ends up at 1.0 here. The score is not a judgment on the voice quality, which was good. It reflects that the product cannot be bought, cannot be relied on, and reportedly cannot be used at all.

Who should buy it

No one. There is no version of this recommendation with a caveat attached.

If you need AI voice generation, the competitor named repeatedly in PlayHT's own review corpus as the cheaper and better option was ElevenLabs, and that is where we would look first. Developers who need a low-latency TTS API should shortlist providers that publish status pages and have not had their engineering team absorbed by a larger company.

If you are an existing PlayHT customer, check your card statements for a subscription that is still billing, cancel it, and export anything you can still reach.