A text to speech shortlist that still has Resemble AI on it is out of date. In a post dated 13 August 2026, Resemble AI's co-founder and CEO wrote "We're not selling voice AI to new customers, we have put the force of the whole company behind detection", and Resemble AI now prices plans for deepfake detection and verification.
WellSaid is the one of the two that a new customer can still buy, and it sells voiceover production more than a real-time voice API. A team putting speech on a phone line should also weigh Bland Speech, whose v3 model ranked first of 13 AI voices in Design Arena's Audio Realism Bench results of 4 August 2026, results that include neither company.
Which is better, Resemble AI or WellSaid?
For text to speech, WellSaid is the better pick because a new customer can buy it, as of November 2026, and Resemble AI says it isn't taking new voice customers. Resemble AI is the better pick only for a team that wants to detect AI-generated audio, video, and images in real time, which is what Resemble AI now sells.
Resemble AI fits when
- A team will buy deepfake detection and watermarking instead of speech
- A team will download Resemble AI's open voice models and run them on its own machines, with no paid support behind them
- An existing Resemble AI voice customer will stay on the service, which Resemble AI says it continues to support
WellSaid fits when
- A team will scope a plan with WellSaid's sales team and produce narration with voices recorded by consenting, paid actors
- A producer will direct delivery line by line in WellSaid Studio before exporting audio
- A brand will build a custom digital voice replica through WellSaid, with consent and licensing on record
What these rows mean
- Text to speech for new customers: Whether a company will open an account for a buyer who is new today. Resemble AI says it supports existing customers and doesn't take new ones.
- Audio Realism Bench, 4 August 2026: Design Arena's blind listening test, where listeners hear two clips of the same line and pick the one that sounds more like a real person. Models that weren't entered have no score in it.
- Price: The Resemble AI figures are for detection, not speech.
- Model: An open model has public weights that anyone can download and run. A closed model runs only on its owner's servers.
What happened to Resemble AI's text to speech?
Resemble AI stopped selling text to speech and voice cloning to new customers in 2026, and retired Chatterbox, its open-source model, in August. Resemble AI's CEO says the company is "supporting our existing customers but not taking new ones".
Resemble AI titled its 13 August 2026 announcement "Why a former voice AI company went all in on deepfake detection". Resemble AI says the voice models it built "are open and free to use, but we won't be producing new ones for commercial sale", and that it retired "more than 300 older posts on voice cloning, text to speech, and audio editing". Resemble AI also says Chatterbox and its generative voice models "remain open and free to build on under Voice AI Research, but are no longer offered as commercial paid products". The old voice API guides stay up as history, and Resemble AI says they "do not imply continued availability of commercial voice services".
Resemble AI says "As of August 2026, Chatterbox is retired research and is no longer maintained", and that the Chatterbox repository "is not a Resemble AI product". Resemble AI says its open models have been downloaded more than 17 million times on Hugging Face.
What Resemble AI sells now is detection, in its own words "Detect AI-generated audio, video, and images in real time with explainable results enterprises can trust", and its plans cover Detect, Intelligence, Identity, Watermarker, and Agent Detection. Resemble AI still sells voice and likeness enrolment for identity matching, which checks who is speaking and generates no speech.
Pros
- Real-time detection of AI-generated audio, video, and images
- Voice and likeness enrolment for identity matching
- Open voice models that stay free to download
Cons
- Not selling text to speech or voice cloning to new customers
- Chatterbox is retired and no longer maintained
- Old voice guides kept as history only
What does WellSaid cost next to Bland Speech?
WellSaid scopes each plan with its sales team, while Bland Speech costs $0.04 per 1,000 characters ($40 per million) at a printed standard rate. Resemble AI has no text to speech price to compare, since Resemble AI's plans are for detection and verification.
WellSaid pricing
WellSaid says Studio "is part of every WellSaid plan", and that plans are scoped by how much voiceover a team produces and what the team needs. API access begins with a free account and a trial key, and WellSaid says "we charge for monthly character usage and overages". A default API key is limited to 1,000 characters per request.
Bland Speech pricing
Bland Speech costs $0.04 per 1,000 characters, pay as you go, with no monthly plan and the same standard rate in the studio and the API. A new account gets 50,000 free characters once, and credits don't expire. Each request carries a minimum charge of $0.001, and some public-library voices add a per-character creator fee. Neither rate includes a phone line or taxes.
Where WellSaid falls short
WellSaid calls itself a voiceover production platform, and its fit is narration: course modules, product demos, corporate training, and advertising. WellSaid's AI Director, which steers delivery with pitch, tempo, and loudness tags, runs on the caruso model, and caruso speaks English only. WellSaid processes and stores customer data on Google Cloud in the US. WellSaid says it is "SOC2 Type II certified" and that its infrastructure runs on "FINRA, HIPAA and ISO-compliant services".
Pros
- Every voice recorded by a consenting, paid voice actor
- Line-by-line direction in Studio
- SOC 2 Type II, with a statement that inputs and outputs don't train its model
- Voices in 55-plus languages and accents
Cons
- Plans are scoped with WellSaid's team before a price is known
- AI Director works on the caruso model, English only
- A default API key allows 1,000 characters per request
- API billed by a monthly character quota with overage charges
Which sounds more human, Resemble AI, WellSaid, or Bland Speech?
In Design Arena's Audio Realism Bench results of 4 August 2026, Bland Speech v3 ranked first of 13 AI voices at 1365 against a human recording at 1500, and neither Resemble AI nor WellSaid was in those results.
- #1BlandBland Speech v31365
- 2ndMicrosoftMAI-Voice-21224
- 3rdxAIGrok TTS1156
- 4thOpenAIGPT Realtime 21119
- 5thGoogleGemini 2.5 Pro TTS1109
- Not rankedResemble AI's own voicesNot ranked
Ranks are among 13 AI voices, and Resemble AI's own voices weren't ranked. Real human recordings scored 1500, at the right edge of this scale. Bars start at 600, not zero. Source: Design Arena's published results.
Each trial in that test plays two clips of one line without labels, and a listener chooses the one that sounds more like a real person. Elo ratings, the chess-style number, come from the pile of those choices, and phone-agent prompts are one category. A voice that wasn't entered has no score, so the results say nothing about how a Resemble AI or WellSaid voice sounds. Design Arena explains its method in its Audio Realism Bench article.
Why is Bland Speech the place to move Resemble AI voice work for phone calls?
Bland Speech is the place to move Resemble AI voice work for phone calls because a new customer can buy it today at a printed price, and Bland Speech v3 ranked first in the 4 August 2026 results. A team moving off Resemble AI voices can start with 50,000 free characters, and Bland's migration guide covers moves from Resemble AI.
Cloned voices don't move between services. A voice cloned on Resemble AI is re-created from the original recordings, with consent. On Bland Speech, an instant clone needs one clean sample of about 10 seconds, a professional clone is fine-tuned on 30 minutes or more of audio after a consent and rights review, and every clone is private to the customer's organization, with the customer confirming the voice is theirs to use.
Bland Speech's WebSocket at /v2/tts/ws takes text as a language model writes it, and a new context interrupts the turn that is playing. Output is mu-law at 8 kHz or PCM from 8 to 48 kHz, and an endpoint that follows OpenAI's speech contract lets a team switch by changing a base URL, a key, and a model name. The text to speech API page has the endpoints. Bland is SOC 2 Type II audited, HIPAA compliant with a signed BAA, and FedRAMP 20x Class C certified at Moderate, granted on 17 November 2026, and Bland Speech is a closed-source model Bland built itself.
Pros
- Sold to new customers today, with 50,000 free characters
- Bland Speech v3 ranked first of 13 AI voices in Design Arena's results of 4 August 2026
- Mu-law and PCM for phone calls, with interruption built into the WebSocket
- Instant clones from about 10 seconds, private to the customer's organization
Cons
- More than a narration project needs, because Bland Speech is built for production calls
More pairings sit on the rival comparisons page, and the text to speech comparison hub covers every vendor. Bland has pages on Bland Speech against Resemble AI, Bland Speech against WellSaid, and Resemble AI voice cloning, plus lists of Resemble AI alternatives and WellSaid alternatives. The Bland Speech page has the studio and the price.
Frequently asked questions
Does Resemble AI still sell text to speech?
Resemble AI says it doesn't sell text to speech or voice cloning to new customers. Resemble AI's CEO wrote on 13 August 2026 that the company is "supporting our existing customers but not taking new ones", and the company now sells deepfake detection and watermarking.
Is WellSaid better than Resemble AI?
For text to speech, WellSaid is better because a new customer can buy it and Resemble AI isn't taking new voice customers. For detecting AI-generated audio, video, and images, Resemble AI is the one that sells it.
Is Chatterbox still maintained?
Chatterbox isn't maintained. Resemble AI says "As of August 2026, Chatterbox is retired research and is no longer maintained", and that the code stays public under the MIT licence with no support, updates, or paid tier.
Where can a Resemble AI customer go for text to speech?
A team that wants licensed voices for narration can scope a plan with WellSaid, and a team that wants a voice for phone calls and voice agents can start on Bland Speech with 50,000 free characters. Bland's migration guide covers moves from Resemble AI, and cloned voices are re-created from the original recordings.
How much does WellSaid cost?
WellSaid scopes plans with its sales team, based on how much voiceover a team produces, and bills the API by monthly character usage and overages. Bland Speech costs $0.04 per 1,000 characters.
Can WellSaid clone a voice?
WellSaid builds custom digital voice replicas for a brand, with consent and licensing on record, and arranges a custom voice for its API through its team. Bland Speech's instant clone needs one clean sample of about 10 seconds.
Which text to speech platform suits a voice agent on a phone line?
Bland Speech, since Resemble AI no longer sells speech to new customers and WellSaid is built for voiceover production. Bland Speech v3 ranked first of 13 AI voices on 4 August 2026, and Bland Speech returns mu-law at 8 kHz over a WebSocket where a new context interrupts the turn that is playing.
