Modulate Raises $25M to Expand Voice Intelligence and AI Audio Detection

Modulate, a Boston-based voice intelligence startup, has raised $25 million in new funding to expand its platform for analyzing, monitoring and securing voice interactions.

The company develops a suite of smaller AI models that can perform transcription, emotional analysis, synthetic voice and deepfake detection, intent analysis and policy enforcement. Its technology is designed for enterprises, particularly organizations operating in regulated industries and businesses deploying voice-based AI agents.

The latest funding was led by Future Ventures, with participation from Hyperplane and Lakestar.

The new financing comes as enterprises increasingly deploy AI-powered voice systems for customer service and other interactions, creating demand for tools that can determine not only what was said, but also how a conversation unfolded, whether an AI-generated voice was involved and whether the interaction complied with company policies.

From gaming voice modulation to enterprise voice intelligence

Modulate was founded in 2017 by Mike Pappas and Carter Huffman, who met while studying physics at MIT.

The company's early focus was voice modulation for gaming. It later developed a voice-based moderation product before shifting toward broader voice intelligence and AI audio analysis as voice AI technology advanced.

Today, Modulate operates more than 100 models across different categories.

The company's technology is designed to extract information from speech and then analyze that information to understand conversations and detect potentially problematic behavior.

The first group consists of signal extraction models. These models examine characteristics such as vocal emotion, tone, language and whether a voice appears to be synthetic.

The second group consists of analysis and detection models. These systems focus on the meaning and intent behind conversations, including what a customer is attempting to communicate, whether a caller may be violating policies or whether an interaction could involve fraudulent activity.

This combination allows Modulate to go beyond conventional speech-to-text systems.

Looking beyond transcription

Transcription has become a standard capability across voice AI platforms. Modulate's founders believe that simply converting speech into text does not capture the full context of a conversation.

Carter Huffman, co-founder of Modulate, told TechCrunch that the company sees an opportunity to understand the nuance of human conversations rather than simply transcribing them.

That distinction becomes particularly important for enterprises handling customer calls.

A transcript can show what a customer said, but it may not fully explain the customer's intent, emotional state or whether the interaction was successful.

For example, a customer may remain polite during a conversation with an AI agent while still being dissatisfied with the outcome.

Modulate's analysis tools are designed to identify those differences.

The company can provide enterprises with more granular information about customer intent and responses, helping organizations evaluate voice interactions beyond basic sentiment classification.

More than 100 specialized models

Rather than relying on a single large model for every task, Modulate has built a collection of more than 100 models.

The models are largely organized around signal extraction and analysis or detection.

This architecture allows the company to add new capabilities as requirements change.

Huffman said the smaller-model approach also reduces the need for specialized hardware and large amounts of computing resources.

That can become increasingly important as the cost of running AI workloads grows.

Smaller specialized models can also be trained for particular tasks and incorporated into the broader system as needed.

An orchestration layer can then determine which models should be used for a particular voice interaction.

For enterprises, this approach can potentially provide a more targeted way to analyze conversations without requiring every task to run through a large general-purpose model.

Deepfake detection and voice security

One of Modulate's key areas is synthetic voice and deepfake detection.

Voice cloning technology has become increasingly accessible, making it possible to create convincing synthetic versions of people's voices.

That creates new risks for organizations receiving calls from customers, employees or other parties.

Modulate can analyze voice interactions and alert organizations, including call centers, to potential scams or synthetic audio.

The company said its technology is also being used to monitor cyberattacks conducted through voice calls.

This puts Modulate at the intersection of voice AI, cybersecurity and fraud prevention.

As enterprises increasingly rely on voice communication, determining whether an interaction is genuine can become as important as understanding the content of the conversation.

Monitoring AI agents

Modulate also provides tools for organizations deploying AI-powered voice agents.

As companies use AI agents to handle customer service calls, they need to evaluate whether those systems are performing appropriately.

That includes assessing the quality of conversations, determining whether customers achieved the desired outcome and checking whether an AI agent followed company policies.

In regulated industries, policy compliance can be especially important.

Modulate's technology can sit alongside an organization's existing voice stack and analyze calls without necessarily replacing the underlying voice infrastructure.

This positioning allows the company to focus on intelligence and monitoring while enterprises continue using their preferred voice platforms.

The company can evaluate AI agent responses and flag interactions that may violate predefined rules.

A growing market for voice AI infrastructure

Modulate's funding comes as the voice AI market expands beyond simple transcription and voice generation.

A growing number of startups are building technologies to make AI voices more natural, while another group is focusing on understanding and securing voice interactions.

Modulate operates in the latter category.

Its products address several overlapping areas, including voice intelligence, customer experience analytics, fraud detection, deepfake identification and AI agent compliance.

The company therefore has an opportunity to serve organizations that may already be investing in voice AI but need additional systems to monitor what happens after a conversation begins.

For call centers and other customer-facing organizations, this can provide another layer of oversight.

Building for privacy and on-device deployment

Modulate currently has approximately 40 to 45 employees.

Following the latest funding, the company plans to add around 10 employees in the coming months, primarily to strengthen its model-building capabilities.

It is also working to expand its on-premises and on-device deployment capabilities.

Privacy is an important consideration for organizations analyzing voice conversations because calls can contain sensitive personal, financial or business information.

On-premises and on-device deployment can give enterprises additional control over where voice data is processed.

For regulated industries, the ability to keep sensitive information within controlled infrastructure can be an important consideration when evaluating AI systems.

Modulate's development of smaller specialized models could support this strategy by making AI workloads more practical to run in environments with tighter computing constraints.

$25M to build the voice intelligence layer

The $25 million funding round gives Modulate additional capital as it expands its model portfolio and works to build a broader intelligence layer around voice interactions.

Founded by Mike Pappas and Carter Huffman, the company has evolved from gaming-focused voice modulation into an enterprise platform covering transcription, voice analysis, deepfake detection, intent recognition and AI agent monitoring.

The company now serves a varied customer base, with particular focus on deepfake detection, call-center monitoring and regulated environments.

Its approach of combining more than 100 specialized models allows Modulate to analyze different dimensions of a conversation rather than relying solely on transcription or broad sentiment scores.

As AI-powered voice agents become more common, enterprises will increasingly need to understand not only what their systems say, but also how customers respond, whether conversations follow rules and whether the voice on the other end is authentic.

Modulate is using its new funding to build around that emerging requirement.