Subtle Computing, a startup based in California, is addressing the challenge of capturing voices clearly in loud settings with its innovative voice isolation models. This technology has the potential to greatly improve voice-based AI products and services.
Voice AI-powered consumer apps are currently experiencing significant growth. AI meeting assistants such as Granola, Fireflies, Fathom, and Read AI have garnered considerable attention from both users and investors. Established companies like OpenAI, ClickUp, and Notion have begun incorporating voice transcription solutions. App developers like Wispr Flow and Willow are actively developing voice dictation tools. Furthermore, hardware companies like Plaud and Sandbar are utilizing devices to transcribe speech and leverage AI for generating insights and facilitating interactions.
One major obstacle for these companies is ensuring reliable voice capture across diverse environments, including noisy locations such as busy cafes or open-plan offices.
To overcome this, Subtle Computing has created a comprehensive voice isolation model capable of accurately understanding speech even in noisy conditions. Chen mentioned that numerous companies are dedicated to voice understanding, but that device manufacturers often send voice data to the cloud for cleaning, which is inefficient.
Instead of using a single model for all devices, the startup customizes models to match the specific acoustics of each device and adapt to the user’s unique voice.
Chen stated, “We’ve discovered that preserving the acoustic characteristics of a device leads to significantly better performance compared to generic solutions. This also enables us to deliver personalized solutions to users.”
Tyler Chen, David Harrison, Savannah Cofer, and Jackie Yang, who met at Stanford, founded the company. Chen, Harrison, and Yang were doctoral candidates, while Cofer was pursuing her MBA. They collaborated in Steve Blank’s Lean Launchpad course, where they explored alternative computing interfaces and laid the groundwork for Subtle Computing.
Techcrunch event
San Francisco
|
October 13-15, 2026
Chen explained, “As our interactions with AI increase, we’re heading towards a future where we communicate with devices using our voices. However, it’s critical to consider how well our devices comprehend us in various daily environments. Voice technology still struggles in noisy coffee shops or shared offices where privacy is a concern.”
The startup claims its voice isolation model, which is just a few megabytes in size and has 100ms of latency, can run on certain devices. For other devices, it can utilize a separate model to transcribe speech into text. Chen noted that the isolation model enhances the accuracy of the transcription model, leading to improved transcript quality.
Subtle Computing announced that Qualcomm has accepted it into its voice and music extension program, meaning its technology will be compatible with Qualcomm chips and available on devices from various OEMs.
The company has secured $6 million in seed funding, led by Entrada Ventures, with contributions from Amplify Partners, Abstract Ventures, and angel investors, including Biz Stone of Twitter, Evan Sharp of Pinterest, and Johnny Ho of Perplexity.
Karen Roter Davis, Managing Partner at Entrada Ventures and former director of an early project at X (Alphabet), pointed out that voice AI is a crowded field, and while voice-based interactions are becoming more common, the overall experience needs improvement. She believes Subtle Computing’s focus on voice isolation offers a unique approach to the market.
Davis stated, “While the impact of AI on our daily time allocation is debatable, advancements in computing power and machine learning/AI create opportunities for significant breakthroughs in voice interface technology, provided they are implemented correctly. Subtle Computing delivers reliable, user-friendly, and enjoyable voice interfaces that excel in both noisy and quiet environments, making it a true game changer.”
The company mentioned partnerships with an unnamed consumer hardware brand and an automotive brand to implement its solutions. However, Subtle Computing aims to be more than just a model provider.
The startup also plans to unveil a consumer product that integrates both hardware and software next year, though specific details remain undisclosed.
