California-based startup Subtle Computing uses a unique voice separation model to tackle the problem of capturing people’s voices in noisy environments. This technology could benefit voice-based AI products and services.
Consumer apps that use voice AI are currently experiencing tremendous growth. Meeting note-taking AIs such as Granola, Fireflies, Fathom, and Read AI are gaining attention from both users and investors. Existing companies like OpenAI, ClickUp, and Notion are integrating audio transcription solutions. App makers like Wispr Flow and Willow are working on voice dictation. Additionally, there are hardware companies like Plaud and Sandbar that use devices as a medium to transcribe audio and use AI to generate insights and interact.
One of the challenges for these companies is collecting user feedback in any environment, such as a noisy cafe or office.
To address this, Subtle Computing has developed an end-to-end speech separation model that can understand what users are saying even in noisy environments. Chen said many companies are working on speech understanding. He pointed out that device manufacturers sometimes send audio to the cloud for a clean output, but that’s not efficient.
Rather than training one model that works across devices, the startup trains specific models to match the acoustics of a specific device and adapt to the user’s voice.
“What we found is that preserving the acoustic properties of a device can result in orders of magnitude better performance than generic solutions. This also means we can provide personalized solutions to users,” said Chen.
The company was founded by Tyler Chen, David Harrison, Savannah Cofer, and Jackie Yang, who met at Stanford University. Chen, Cofer, and Yang were pursuing Ph.D.s, and Harrison was pursuing an MBA. They gathered at Steve Blank’s Lean Launchpad course, worked on alternative interfaces for computing, and began building Subtle Computing.
tech crunch event
san francisco
|
October 13-15, 2026
“As we interact more with AI, we are moving toward a future where we talk to our devices,” Chen said. “But the obvious question is how well do our devices understand us as users in all the environments we work in every day? Whether it’s a very noisy coffee shop or a shared office with other people around, we might be talking about private things. Voice doesn’t work that way today,” he added.
The company said it can run this model for audio separation only on some devices. This model is only a few megabytes in size and has a latency of 100 milliseconds. The company can also run separate models to transcribe audio and provide text output to other devices. Chen said that the separation model allows the company’s transcription model to better understand users and, as a result, create more accurate transcriptions.
Subtle Computing said Qualcomm selected the startup as a member of its Voice and Music Expansion Program. This means the startup’s technology is compatible with Qualcomm’s chips and can be used in devices manufactured by OEMs.
The company has raised $6 million in seed funding led by Entrada Ventures with participation from Amplify Partners, Abstract Ventures, and angel investors including founders such as Twitter’s Biz Stone, Pinterest’s Evan Sharp, and Perplexity’s Johnny Ho.
Karen Roter Davis, managing partner at Entrada Ventures and former director of early projects at She believes the startup’s focus on audio separation will bring a different perspective to the market.
“While we can debate whether AI will increase or decrease the amount of time spent on a daily basis, we can all agree that advances in computing power and machine learning/AI present an opportunity for breakthroughs in voice interfaces, if executed properly,” Davis said. “Subtle Computing meets people wherever they are and provides a reliable, easy, and enjoyable audio experience with a voice interface that can withstand both extreme noise and extreme silence. It’s a game-changer,” she added.
The company said it is also partnering with consumer hardware and automotive brands (which it did not name) to roll out its solutions. But Subtle Computing doesn’t want to be just a model supplier to other companies.
The company also said it plans to introduce consumer products spanning both hardware and software next year, but did not provide details.
