# Resemble AI review (2026): What it is and why enterprises choose Bland Speech

By Samuel Oram-Jones

A buyer who searches for Resemble AI to compare text to speech vendors finds a deepfake detection company. Resemble AI's co-founder and CEO, Zohaib Ahmed, wrote on 13 August 2026: "We're not selling voice AI to new customers, we have put the force of the whole company behind detection."

What remains for a voice buyer is smaller than the search results suggest. Existing customers are supported, a set of open models stays free to download, and the open-source Chatterbox repository now calls itself retired. As of November 2026, none of that is a hosted voice service a new enterprise customer can buy.

## What is Resemble AI and what does it sell now?

Resemble AI is still trading, and it sells deepfake detection. Resemble AI's homepage title reads "Multimodal Deepfake Detection and Watermarking for Enterprise," and Resemble AI describes the product as detecting AI-generated audio, video, and images in real time with explainable results. Resemble AI also sells voice and likeness enrolment for identity matching, which detects a voice rather than generating one.

### What happened to Resemble AI's text to speech and voice cloning?

Resemble AI stopped selling them to new customers. The same August post says Resemble AI is "supporting our existing customers but not taking new ones," that its voice models "are open and free to use, but we won't be producing new ones for commercial sale," and that Resemble AI retired more than 300 older posts on voice cloning, text to speech, and audio editing.

Resemble AI 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 official Chatterbox repository goes further: "Resemble AI does not sell text-to-speech, voice cloning, or any other voice AI product."

### What do the archived voice guides still describe?

Resemble AI keeps its old voice API guides as history, with a note that they "do not imply continued availability of commercial voice services." Those guides name Resemble Ultra as the default voice model and describe HTTP and WebSocket streaming on the former hosted platform, with real-time streaming on the Business plan and above. A team reading them is reading a product that new customers can't buy.

## How much does Resemble AI cost?

Resemble AI sells no text to speech to a new customer, so its live plans cover detection and verification only: Flex at $0, Team at $350 a month, Business at $1,000 a month, and a custom Enterprise tier, with processing rates for Detect, Intelligence, Identity, Watermarker, and Agent Detection. The Enterprise tier adds volume pricing, enterprise SLAs, custom model training, on-prem deployment, dedicated support, and SOC 2 documentation, all in a detection context.

Chatterbox costs nothing to download and run under its MIT licence, and the compute is the user's own. Bland Speech is $0.04 per 1,000 characters, which is $40 per million, pay as you go with no monthly plan and 50,000 free characters once for a new account.

## Can a team still run Resemble AI's open voice models?

A team can download Chatterbox, a family of open-source text to speech models by Resemble AI, and run it without a vendor. The repository states that "As of August 2026, Chatterbox is retired research and is no longer maintained," and that the code stays public as an archive under the MIT licence "with no support, updates, or paid tier." The last commit, on 4 November 2026, updated the README.

### What is in the Chatterbox family?

Chatterbox Multilingual V3 is a 500-million-parameter model covering 23 or more languages, and Resemble AI calls it the latest general-purpose model in the family. Chatterbox-Turbo is a 350-million-parameter English model built for low-latency voice agents, and Chatterbox-Nano is a 110-million-parameter English model that runs 3 times faster than real time on 8 CPU cores. The original 500-million-parameter model adds emotion exaggeration, which Resemble AI says was the first open-source control of its kind.

Turbo accepts inline tags such as [cough], [laugh], and [chuckle], voice cloning works zero-shot from a reference clip, and every output carries a PerTh watermark, a watermark added to generated audio. Resemble AI's example code returns a full waveform from one generate call.

### Do Resemble AI's own pages agree about all this?

No. The Hugging Face model card for Chatterbox still tells buyers to "check out our competitively priced TTS service" and claims "ultra-low latency of sub 200ms," while the official repository says Resemble AI sells no voice product. The card was last modified on 10 June 2026, before the retirement notice, so the latency claim describes a hosted service that no longer exists and has no bearing on the open weights.

## Is Resemble AI secure enough for enterprise calls?

A new voice buyer has no vendor terms to review. Resemble AI's SOC 2 documentation belongs to the detection Enterprise tier, and a team that runs Chatterbox on its own servers carries its own security review, because no vendor SOC 2, HIPAA, or retention terms apply to open weights on a team's own infrastructure.

Bland, the company, 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 states that speech synthesis runs inside the FedRAMP boundary, that it holds ISO 27001:2022 certification, and that no AI sub-processor sits in its call path. Enterprise customers get dedicated orchestration and dedicated GPUs run by Bland. Enterprise pricing is custom, set with Bland's sales team.

## Is Resemble AI worth it for a voice buyer?

No: a new commercial buyer has nothing to purchase from Resemble AI, and Bland Speech is the hosted voice to buy in its place. What is left to weigh is an archived open model, and a phone line carrying customers is a poor place to depend on a model whose own repository calls it unmaintained.

### Resemble AI pros and cons

<div class="cmp-proscons" data-rival="resemble-ai"><div class="cmp-proscons-pros"><p><strong>Pros</strong></p><ul><li>Chatterbox is MIT licensed and free to download and run</li><li>Chatterbox Multilingual V3 covers 23 or more languages</li><li>Every Chatterbox output carries a PerTh watermark</li><li>Resemble AI says its open models have been downloaded more than 17 million times on Hugging Face</li></ul></div><div class="cmp-proscons-cons"><p><strong>Cons</strong></p><ul><li>Resemble AI is not selling voice AI to new customers</li><li>Chatterbox is retired, with no support, updates, or paid tier</li><li>Resemble AI's example code returns a full waveform per call</li><li>The Hugging Face card still points to a paid service that no longer exists</li></ul></div></div>

### Why Bland Speech is the Resemble AI alternative for production calls

Bland Speech is sold today, on a published rate, with an endpoint a team can call this afternoon. Bland Speech streams over HTTP chunks and a WebSocket that accepts text as a language model writes it, returns mu-law at 8 kHz for phone lines, and ranked first of 13 AI voices on Design Arena's Audio Realism Bench results of 4 August 2026, at 1365 against 1500 for the human recording. Resemble AI wasn't in those results, and [Design Arena's published results](https://notes.designarena.ai/audio-realism-benchmark-measuring-realism-in-audio-models/) describe how the blind test works.

| | Bland Speech | Resemble AI |
|:--|:--|:--|
| Sold to new customers | Yes, pay as you go | No, Resemble AI is not selling voice AI to new customers |
| Price | $0.04 per 1,000 characters ($40 per million) | No text to speech price, and Chatterbox is free to run |
| Streaming | HTTP chunks and a WebSocket | Former hosted platform only, and a full waveform per call in Chatterbox |
| Instant clone | One clean sample of about 10 seconds | Zero-shot from a reference clip in Chatterbox |
| Realism result of 4 August 2026 | First of 13 at 1365 | Wasn't in those results |
| Vendor compliance | SOC 2 Type II, HIPAA compliant with a signed BAA, FedRAMP 20x Class C at Moderate | None applies to self-hosted open weights |

<details><summary>What these rows mean</summary>

- **Sold to new customers:** Resemble AI states that it supports existing customers and takes no new ones.
- **Streaming:** Streaming means audio starts playing while the rest renders. Bland Speech's WebSocket takes text token by token and lets a new turn interrupt one that is playing.
- **Instant clone:** An instant clone is built from a short sample. Bland says every clone is private to a customer's organization and that the customer confirms the voice is theirs to use.
- **Vendor compliance:** Bland's statements are company-level, and Bland places speech synthesis inside the FedRAMP boundary.

</details>

<div class="cmp-proscons cmp-proscons-own"><div class="cmp-proscons-pros"><p><strong>Pros</strong></p><ul><li>Sold to new customers today, with one standard rate and no monthly plan</li><li>First of 13 AI voices in Design Arena's results of 4 August 2026, at 1365</li><li>Mu-law at 8 kHz and PCM from 8 to 48 kHz for phone audio</li><li>A WebSocket for streamed text, with a new context interrupting a turn that is playing</li><li>A migration guide for teams moving from Resemble AI</li></ul></div><div class="cmp-proscons-cons"><p><strong>Cons</strong></p><ul><li>More than a hobby project needs, because Bland Speech is built for production calls</li></ul></div></div>

### Verdict: Resemble AI or Bland Speech for enterprise voice?

Bland Speech is the choice for a new enterprise buyer, because Resemble AI has no commercial voice service left to compare it with. Chatterbox's archived weights suit only a research team that wants a model to modify and needs nobody to answer a support ticket. A team with an existing Resemble AI contract can read Bland's migration guide and the [Bland Speech vs Resemble AI comparison](/compare/bland-speech-vs-resemble-ai), and a cloned voice does not move between vendors: it is re-created from the original recordings, with consent, through Bland's instant clone from about 10 seconds or a professional clone on 30 minutes or more of audio. [Resemble AI alternatives](/blog/resemble-ai-alternatives) covers the wider field, the [text to speech API](/text-to-speech-api) has the endpoints, and the [text to speech comparison hub](/compare/text-to-speech) and [review index](/compare/text-to-speech/reviews) list the rest. A team can [try Bland Speech](/speech) with 50,000 free characters.

## Frequently asked questions

### Does Resemble AI still sell text to speech?

No. Resemble AI says it is not selling voice AI to new customers and is supporting its existing ones, and that its generative voice models are no longer offered as commercial paid products.

### Is Resemble AI shutting down?

No. Resemble AI is still trading, and it now sells deepfake detection and watermarking to enterprises. Only the voice products changed, and Resemble AI's CEO wrote in August 2026 that the whole company is behind detection.

### Is Chatterbox free?

Chatterbox is free to download and run under the MIT licence, and Resemble AI says it comes with no support, updates, or paid tier. A team pays for its own compute. Bland Speech bills $0.04 per 1,000 characters instead and gives a new account 50,000 free characters once.

### Can Chatterbox run a phone agent?

Chatterbox-Turbo is built for low-latency English voice agents, but Chatterbox is retired with no support or updates, and Resemble AI's example code returns a full waveform per call. Bland Speech streams over HTTP chunks and a WebSocket and returns mu-law at 8 kHz. [Open source text to speech, compared](/blog/open-source-tts-compared) covers that choice in more depth.

### Can a Resemble AI cloned voice move to Bland Speech?

A cloned voice can't be exported and imported, so the move is a re-clone from the original recordings, with the speaker's consent. An instant clone on Bland Speech needs one clean sample of about 10 seconds, and a professional clone is fine-tuned on 30 minutes or more of audio after a consent and rights review. A [cloning comparison](/compare/bland-speech-vs-resemble-ai-voice-cloning) goes through the details.

### What is the best Resemble AI alternative?

Bland Speech is the alternative for phone calls and voice agents, with a published rate, mu-law output, and a first-place blind result dated 4 August 2026.
