Resemble AI 2026: Deepfake Detection, Pricing and Voice Cloning

Updated October 2026: Resemble AI no longer sells voice generation to new customers. This review now covers what the company sells today, what happened to its voice cloning and text-to-speech products, and where voice teams can go instead.

If you last looked at Resemble AI for voice cloning, the company you remember isn't the one on resemble.ai today. In August 2026, co-founder and CEO Zohaib Ahmed announced that Resemble had stopped selling voice AI to new customers and put the whole company behind deepfake detection. The homepage headline now reads "Deepfakes are everywhere. So are we."

This review covers what Resemble AI does in 2026, what it costs, how it fits a contact center that runs voice agents, and what to use if you came here for text-to-speech (TTS) or voice cloning.

What Is Resemble AI?

Resemble AI is a deepfake detection and content provenance company for enterprises. It sells APIs and ready-to-use products that tell you whether audio, video, or an image is real, who is speaking, and where a piece of media came from.

The current product line:

• Detect. Real-time detection of AI-generated audio, video, and images. Resemble reports 99.5% accuracy on audio (Podonos benchmark, August 2026), with video at 98.2% and images at 95.8% on internal tests while external testing is in progress.
• Intelligence. Structured insights and a plain-language explanation of each verdict, so a fraud analyst can see why a file was flagged.
• Watermarker. Imperceptible watermarks (the PerTh models) for audio, video, and images, plus C2PA provenance records.
• Identity. Enrolls a speaker from 4 seconds of audio, then verifies or flags that voice on later calls and recordings.
• Meetings. Detection of synthetic voices and faces inside Zoom, Microsoft Teams, Google Meet, and Webex.
• Agent Detection. A model that spots AI agents, billed per session.
• Chrome extension. Deepfake checks in the browser.

Logos on the homepage include Netflix, Paramount, Deutsche Telekom, Telnyx, Five9, 8x8, and Snowflake. On security, Resemble lists GDPR, ISO 27001, SOC 2 Type II, and HIPAA-aligned data handling, with cloud, on-premises, and air-gapped deployment options.

What Happened to Resemble's Voice Cloning and Text-to-Speech?

Resemble built its name on voice cloning and TTS. In an August 13, 2026 post, Ahmed explained the change: Resemble's H1 2026 Deepfake Threat Report counted at least 15,736 people harmed across 821 documented attacks in six months, and the company decided detection was where it should spend all of its effort.

What that means in practice:

• New customers can't buy voice generation. Resemble isn't selling voice AI to new customers and won't produce new voice models for commercial sale.
• Existing customers are still supported. If your team already runs on Resemble voices, talk to your account contact about your contract and roadmap.
• The open models stay open. Chatterbox, DramaBox, and Resemble's other generative voice models live on under Voice AI Research and remain free to build on.
• Old content is gone. Resemble retired more than 300 older posts about voice cloning, TTS, and audio editing, and most now redirect to the announcement.

So if a search result, comparison table, or older review (including earlier versions of this one) quotes Resemble's TTS pricing or voice cloning tiers, treat it as history.

Resemble AI Pricing 2026

Resemble's pricing is now built around detection. There are four plans:

• Flex: $0 per month with no subscription fee. You buy credits as you go. One seat.
• Team: $350 per month, or $280 per month billed annually. Five seats and lower usage rates.
• Business: $1,000 per month, or $800 per month billed annually. Twenty seats.
• Enterprise: custom pricing through sales.

Usage rates on Flex, then Team and Business:

• Detect (audio): $0.035 per second on Flex, $0.015 per second on Team and Business
• Detect (images): $0.035 per image on Flex, $0.015 per image on Team and Business
• Detect (video): $0.07 per second on Flex, $0.03 per second on Team and Business
• Intelligence: $0.025 per second on Flex, $0.015 per second on Team and Business
• Identity search: $0.0005 per call
• Watermarker: $0.0005 per encode, $0.0002 per decode
• Agent Detection: $1 per 1,000 sessions

Volume pricing is available on every line. These are the rates listed on resemble.ai/pricing in October 2026, so check the page before you budget.

Where the math lands

Audio detection is billed per second, so call length drives cost. Screening a full five-minute call (300 seconds) costs about $4.50 on Team or Business and $10.50 on Flex. Across thousands of inbound calls a day, that adds up quickly.

Most of the risk in a support call sits in a few moments: the caller proving who they are, a password reset, a change of payout details. Decide where in the call you need a verdict before you size the contract. Identity search at $0.0005 per call is a different budget line from second-by-second detection.

Resemble AI for Contact Centers and Voice Agents

This is where Resemble's new focus meets customer support. Two of its solution pages target voice teams directly.

Contact center fraud. Resemble analyzes inbound calls in real time and flags AI-generated voices before an agent shares account details. It also surfaces social engineering patterns like authority spoofing and urgency pressure, without transcribing or recording the call. Resemble quotes detection in under 300ms, validation against 250+ generative models across 54 languages, and a Zero Retention Mode that discards audio after analysis. It connects through Genesys Cloud, Five9, 8x8, Microsoft Teams, Zoom, and Vapi, or over SIPREC, SIP trunk, and a direct API.

Voice agents. As more companies put AI agents on the phone, the agent itself becomes something attackers copy. Resemble's pitch is to verify who is calling, watermark your own agent's audio so a cloned copy can be caught, and detect synthetic callers trying to get past verification. Telnyx, for example, built Resemble's watermarking and detection into its own voice agents.

For a support leader, the takeaway is simple. Resemble no longer supplies the voice your agent speaks with. It's a candidate for the security layer around the line.

Voice Cloning and TTS Alternatives to Resemble AI

If you came to Resemble for voice generation, these are the places voice teams are looking now:

• ElevenLabs. The best-known name in expressive TTS and voice cloning, with instant and professional cloning and a large developer ecosystem.
• Cartesia. Low-latency TTS built for real-time voice, with multilingual voices across its Sonic models.
• Rime. TTS voices tuned for conversational, phone-style speech.
• Google, Amazon, and Microsoft. Cloud TTS (including Google Chirp and Amazon Polly) that fits teams already standardized on one of those clouds.
• Chatterbox. If you want to self-host, Resemble's open Chatterbox models are still free to build on.

For a builder-side walkthrough of wiring a TTS provider into an agent, see build a custom voice AI agent with ElevenLabs API.

Where Voiceflow Fits

Voice generation, detection, and the agent are three different layers. Voiceflow is the agent layer: the system that answers the call, works out what the customer needs, takes action in your backend, and hands off to a person when it should.

For customer support teams running voice, that looks like:

Resemble isn't a native Voiceflow integration. Its contact center detection works at the telephony and CCaaS layer (SIPREC, SIP trunk, or the platforms listed above), so it can sit alongside a Voiceflow agent on the same line.

If you're planning AI phone agents, an AI call center, or broader contact center automation, start with the agent and the handoff design, then add the detection and voice layers that match your risk and your brand. Book a demo to see a Voiceflow voice agent on a live line.

Resemble AI FAQ

Does Resemble AI still offer text-to-speech and voice cloning?

Not to new customers. In August 2026 Resemble stopped selling voice AI to new customers and stopped producing new voice models for commercial sale. Existing customers are still supported, and the open-source Chatterbox and DramaBox models remain free to use.

What does Resemble AI do now?

Resemble AI is a deepfake detection company. Its products detect AI-generated audio, video, and images (Detect), explain each verdict (Intelligence), watermark media (Watermarker), verify a speaker's voice (Identity), and protect live meetings and contact center calls.

How much does Resemble AI cost in 2026?

Flex is $0 per month with pay-as-you-go credits. Team is $350 per month ($280 billed annually) and Business is $1,000 per month ($800 billed annually). Audio detection runs $0.035 per second on Flex and $0.015 per second on Team and Business. Enterprise is custom.

Is Resemble AI free?

There's no free tier for production use, but the Flex plan has no subscription fee. You pay only for the detection, watermarking, or identity checks you run.

Can Resemble AI detect deepfakes?

Yes. Detection is now the company's core product. Resemble reports 99.5% accuracy on audio deepfakes in the Podonos benchmark (August 2026), and it covers video and images as well.

What are the best Resemble AI alternatives for voice cloning?

ElevenLabs is the most common replacement for voice cloning and expressive TTS. Cartesia and Rime are strong picks for real-time phone conversations, and Google, Amazon, and Microsoft offer cloud TTS. All of these work as voice output in a Voiceflow agent.