Modulate raises $25M to expand its audio-native AI tools for developers
The AMW Read
The $25 million round and company-reported audio volume update a specialist model provider's position, with the clearest implications in audio analysis.
Modulate raises $25M to expand its audio-native AI tools for developers
Modulate has raised $25 million in a round led by Future Ventures, with returning investors Hyperplane and Lakestar participating. The company plans new developer SDKs and APIs, industry-specific models, partner integrations, and more deployment options. Its Velma platform analyzes raw audio for signals including emotion, intent, harassment, and possible voice deepfakes. Modulate says its models process more than 10 million hours of audio each month; that figure is company-reported.
The funding puts a specialist audio-model provider on the map alongside companies building voice generation and voice agents. Modulate addresses a different part of that stack: detecting signals that a transcript can miss while a conversation is in progress. Its uses span gaming moderation, screening callers who may be impersonating health care staff, and assessing voice-agent performance. The company says Velma selects from more than 100 specialized models for a given task and is up to 1,000 times more efficient than using one large model for the same audio. That efficiency claim needs evaluation in each deployment setting.
For builders, the immediate question is whether Modulate's audio analysis improves decisions in their own calls at an acceptable latency and error rate. The company reports 98.9% accuracy for Velma Deepfake Detect on public benchmark data, but a buyer screening staff impersonation should also test real call conditions and the cost of false alarms. Investors can watch whether the planned APIs turn Modulate's reported audio volume into broader developer adoption beyond its original gaming market.