A request to PlayHT's API address in November 2026 never reaches PlayHT. A DNS lookup for the address returns NXDOMAIN, which means the name doesn't exist, and curl stops with "Could not resolve host" before any request is sent.
A migration normally starts by exporting from the old provider and running both side by side, and neither is on offer here. What a team has is its own code, the audio it saved, and PlayHT's developer docs, which are still online. That is enough to map each call onto Bland Speech and rebuild.
Is PlayHT still working?
As of November 2026, PlayHT's marketing site and API address don't resolve, its web app fails with an expired TLS certificate, and no PlayHT page announces an end date or says why.
PlayHT's developer docs still load and still describe the product, with pages marked "Updated about 1 year ago" and a models page last changed on 15 September 2025. A lookalike domain spelled without the dot between the two words is a parking page, not PlayHT's site, which matters to a developer correcting what looks like a typo.
Everything in PlayHT's docs therefore describes a 2025 service that a buyer can't reach today, and those pages serve only to map old calls onto new ones.
What maps to what when moving from PlayHT to Bland Speech?
PlayHT's HTTP streaming maps to Bland's POST /v2/tts, its WebSocket API with input streaming maps to Bland's WSS /v2/tts/ws, and its Node.js and Python SDKs map to plain HTTP and WebSocket calls.
What these rows mean
- Text from a language model: Bland describes the PlayHT job as replacing a generated socket URL and request-level speech commands. Bland's socket is configured by one
initmessage and carries text deltas inspeakmessages, withend_of_turnclosing each turn. - Models and voices: Bland's voice decides its model, so there is no model field to map, and
BTTS_V3voices are the current generation. PlayHT rated emotion as better on Play 2.0, good on Play 3.0 Mini, and best on PlayDialog. - Speed: each figure is the company's own. PlayHT's three come from a table in docs updated in 2025, and Play 3.0 Mini's 190ms is below Bland's 199ms.
- Client code: Bland's
bland-ttsNode package calls Bland's older/v1/speakendpoint, and new integrations call/v2/ttsor the WebSocket directly. - Phone audio: Bland's mu-law is fixed at 8,000 Hz. PlayHT published a guide to streaming audio into a phone call through Twilio.
- Price: PlayHT's pricing site is the address that doesn't resolve.
Two parts of a PlayHT integration have no like-for-like partner. One is model choice: a team that picked PlayDialog for emotive speech or Play 3.0 Mini for speed picks nothing on Bland, because the voice it passes decides the model, and Bland says to prefer a BTTS_V3 voice. The other is delivery: Bland publishes a verbatim tag set such as [say warmly] and [laughs], placed before the words they shape, two controls called expressiveness and stability from 0.0 to 1.0, and pause markers such as <|0.8|>, and a team rewrites whatever it used on PlayHT into those.
What are the steps to switch from PlayHT to Bland Speech?
A switch takes six steps: gather what is still local, build a voice table, replace SDK calls, request phone audio directly, rewrite delivery, and test against saved audio.
- A team gathers what it still holds: its integration code, saved PlayHT audio, and the original recordings behind any cloned voice. An API that doesn't resolve can't be asked for anything, so those local files are the inputs.
- A team builds a voice table with each PlayHT voice in one column and a Bland voice ID beside it. A GET to
/v1/voicesreturns every usable voice, and three built-in voices, Karen, Matthew, and River, are open to every organization. - A team replaces SDK calls with HTTP or WebSocket calls. A voice agent opens one socket per conversation, sends
initonce, waits forready, and then sends text as a language model produces it:
{ "type": "init", "voice": "29158307-9893-4149-8a75-bc9ce313d64e", "audio": { "encoding": "pcm_s16le", "sample_rate": 48000 } }
{ "type": "speak", "context_id": "turn-42", "text": "Hello" }
{ "type": "speak", "context_id": "turn-42", "text": ", how can I help?" }
{ "type": "end_of_turn", "context_id": "turn-42" }
- A team that sends audio to a phone line asks Bland for telephony audio directly,
{"audio": {"encoding": "mulaw", "sample_rate": 8000}}, instead of converting afterwards. Mu-law is fixed at 8,000 Hz, and any other rate returnsunsupported_sample_rate. - A team rewrites delivery as tags, pause markers, and the two controls, and renders each line in Speech Studio, which uses the same models as the API, before putting it in code.
- A team tests against the audio it saved, because no live PlayHT path exists to run beside Bland. Twenty of the lines callers hear most often, rendered on Bland and played next to saved PlayHT audio for a listener who can't see the labels, show the difference quickly.
A voice agent that leaves one large speak for the end of a turn gets no audio until end_of_turn, because release is driven by text continuing to arrive. Sent as deltas, the same words start playing about a second in, and Bland publishes 199ms to first audio on a streamed request.
Which text to speech API fits a former PlayHT workload?
A former PlayHT workload that ends on a phone line moves to Bland Speech without an audio conversion step, because Bland returns mu-law at 8 kHz directly, and Bland Speech v3 ranked first of 13 AI voices in Design Arena's Audio Realism Bench results of 4 August 2026, at 1365, with the human recording at 1500.
PlayHT wasn't in those results, so no score from that benchmark places PlayHT's voices beside Bland's. The benchmark has native speakers pick, blind, which of two clips of the same text sounds more like a real person, and it includes a Phone Agents category.
Bland owns its speech model and says it has "Zero AI sub-processors in the call path", which a team moving off a provider whose own domain stopped resolving can fairly weigh. Bland is SOC 2 Type II audited, HIPAA compliant with a signed BAA, and FedRAMP 20x Class C certified at Moderate, granted 17 November 2026.
What does the move cost, and what can't come along?
Bland Speech costs $0.04 per 1,000 characters ($40 per million), pay as you go with no monthly plan, and 50,000 characters are free once for a new account. A 10-minute narration comes to $0.36, and each request carries a $0.001 minimum. Enterprise pricing is custom, set with Bland's sales team.
PlayHT's price can't be set beside that, because PlayHT's pricing site is the address that doesn't resolve. A team with an old PlayHT invoice divides it by the characters it sent and compares the result with $0.04.
A cloned voice doesn't come along. Bland says voice IDs are provider-specific and can't be reused, so each voice is cloned again on Bland from the original recording, with the speaker's permission. An instant clone needs one clean sample of about 10 seconds, a professional clone is fine-tuned on 30 minutes or more of audio and goes through Bland's consent and rights review, and every clone is private to the customer's organization.
The PlayHT alternatives list compares the other places a team can go, and moving off Hume Octave follows the same plan for another provider whose product has gone quiet. The other text to speech migration guides cover providers that still answer, the text to speech comparison hub holds every comparison, the text to speech API page lists each endpoint, and Bland Speech has the studio.
Frequently asked questions
Has PlayHT shut down?
PlayHT's marketing site and API address don't resolve, its web app fails on an expired TLS certificate, and its developer docs still show 2025 pages. No PlayHT page says why or gives an end date.
Can a PlayHT cloned voice move to Bland Speech?
No. Bland says voice IDs are provider-specific, so a team clones the voice again on Bland from its original recording. PlayHT described instant cloning from 30 seconds of speech, and Bland's instant clone needs one clean sample of about 10 seconds.
Which Bland endpoint replaces PlayHT's WebSocket API?
WSS /v2/tts/ws. A team opens one socket per conversation, sends init once, waits for ready, sends text deltas under one context_id per turn, and sends end_of_turn when the language model finishes.
Do PlayHT's models map to a Bland model?
No model field exists to map. A Bland voice decides its model, btts-3 is the current one, and the x-model response header says which model a voice resolved to.
Can a Pipecat or LiveKit agent use Bland Speech?
Yes. Pipecat supports Bland as a text to speech service the Pipecat team maintains, installed with uv add "pipecat-ai[bland]", and LiveKit Agents has a Bland plugin for Python, installed with uv add "livekit-agents[bland]~=1.8".
