What happened
Modulate secured a $25 million Series B funding round, led by Future Ventures with participation from Hyperplane and returning investor Lakestar. The capital will support AI/ML research, product development, developer relations, and partnerships for its Velma platform, a general‑purpose voice‑understanding engine that analyzes raw audio to extract signals such as emotion, tone, intent, synthetic speech, and conversational behavior. The company reports processing over 10 million hours of audio per month and more than 600 million hours in total, with its transcription model ranking first on Hugging Face’s Open ASR Leaderboard and its deep‑fake detection model achieving 98.9 % accuracy on public benchmarks.
Modulate announced a $25 million Series B round on September 29, 2026. Future Ventures led the round, with Hyperplane and Lakestar also participating. Lakestar had previously led Modulate’s $30 million Series A in 2022, bringing total funding to $60 million according to TechCrunch.
The funding will be allocated to AI/ML research, product and engineering efforts, developer relations, and strategic partnerships. CEO Carter Huffman emphasized that the capital will allow Modulate to expand its team and accelerate the delivery of its voice‑understanding infrastructure to developers building new voice experiences.
Modulate’s Velma platform processes raw audio to extract a suite of signals—emotion, tone, intent, emphasis, synthetic‑speech detection, and conversational behavior—before any transcription occurs. These signals are combined to flag higher‑level events such as fraud attempts, AI‑agent failures, harassment, customer dissatisfaction, and policy violations, enabling real‑time intervention.
The underlying Listening Model (ELM) architecture coordinates more than 100 specialized audio models, which Modulate claims delivers up to 1,000× greater computational efficiency compared with a single large . The company reports analyzing over 10 million hours of audio each month and having processed more than 600 million hours in total.
Modulate’s transcription model currently holds the top spot on Hugging Face’s Open ASR Leaderboard, while its deep‑fake speech detection model leads the same platform’s deep‑fake with 98.9 % accuracy on public data.
Source details: en.wowtale.net
Why it matters
Modulate’s funding underscores growing investor confidence in voice‑centric AI beyond speech generation. By focusing on audio‑level understanding rather than text‑first pipelines, Modulate aims to enable real‑time detection of fraud, harassment, policy violations, and synthetic‑voice attacks across industries such as finance, healthcare, and social media. Its Listening Model (ELM) architecture, which orchestrates more than 100 specialized audio models, claims up to 1,000× efficiency gains versus a single large , potentially lowering the cost of deploying sophisticated voice analytics (estimated at 2.5–6 cents per hour versus Deepgram’s 31–55 cents). If the technology lives up to its benchmarks, it could become a critical layer for AI agents, security teams, and customer‑experience platforms that need to react to vocal cues before transcription, expanding the market for voice‑understanding services.
The investment highlights a shift in AI focus from voice generation to voice comprehension, a segment that remains under‑served by major cloud providers. By extracting meaning directly from audio, Modulate can detect malicious or synthetic content earlier than text‑based systems, which is crucial for fraud prevention, child‑safety, and compliance monitoring.
Modulate’s claimed efficiency advantage could dramatically reduce the cost of large‑scale audio analytics, making sophisticated voice‑understanding accessible to smaller developers and enterprises that previously could not afford high‑priced services like Deepgram.
The platform’s ability to flag policy violations and harassment in real time offers a tangible tool for social‑media platforms and call‑center operators seeking to mitigate abuse, aligning with increasing regulatory scrutiny on AI‑driven content moderation.
If Modulate’s benchmarks hold up under broader, real‑world testing, the company could set new standards for audio‑level AI, influencing how future voice assistants and AI agents interpret human cues, potentially reshaping user experience design across the industry.
The competitive landscape includes general‑purpose models such as OpenAI’s GPT‑4o Realtime and Google Gemini’s Affective Dialog, which aim to incorporate non‑verbal cues. Modulate’s specialized, multi‑model approach provides a contrasting strategy that may appeal to customers prioritizing efficiency and domain‑specific accuracy over a single monolithic model.
Interactive Mechanism: How It Actually Works
Explore the underlying technology behind this development interactively.
crm_get_transaction(id='4092').Which component of an AI application is the machine-learning model itself?
What to watch next
Future developments to monitor include: (1) rollout of Velma’s APIs to third‑party developers and any pricing tiers disclosed; (2) adoption by large enterprises in fraud‑prevention, healthcare, and social‑media moderation; (3) competitive responses from general‑purpose models like OpenAI’s GPT‑4o Realtime and Google Gemini’s Affective Dialog; and (4) independent validation of Modulate’s efficiency claims and performance as the company scales its audio processing volume.
Product rollout: Whether Modulate releases public APIs, SDKs, or pricing tiers, and how quickly developers can integrate Velma into existing voice pipelines.
Enterprise adoption: Announcements of partnerships or contracts with banks, healthcare providers, or social‑media platforms that would validate the platform’s fraud‑detection and moderation capabilities.
Competitive response: Updates from OpenAI, Google, and Deepgram that could either reinforce Modulate’s niche or erode its advantage through similar audio‑understanding features.
verification: Independent third‑party evaluations of Modulate’s transcription accuracy and deep‑fake detection rates, as well as real‑world efficiency measurements against competing solutions.
Regulatory impact: Potential involvement in policy discussions or compliance frameworks, especially as governments scrutinize synthetic‑voice attacks and AI‑driven harassment.