Resemble AI, a San Francisco-based voice artificial intelligence startup, said on Wednesday it had raised $8 million in a funding round led by venture capital firm Javelin Ventures. It added that Comcast Ventures (CMCSA.O) and existing investors Craft Ventures and Ubiquity Ventures also participated in the Series A round. The company said it was using the proceeds to develop further its product to detect if audio is artificially generated, known as a deep fake, with up to 98% accuracy.
The company’s new offering is Resemble Detect, which analyzes audio in real-time to determine whether it is AI-generated or has been manipulated. It works across all media types, including video and audio, and can compare audio against a database of more than 35 years of recorded human speech and language. Several media companies have already deployed the technology, the company said.
Founded in 2019, Resemble AI provides its customers with generative AI software to clone voices, generate personalized messages from celebrities, or create unique real-time conversational agents. Brands, media companies, and gaming studios use its technology. It has more than 200 business customers.
Ahmed and his co-founder Saqib Muhammad founded the company in 2019 after noticing that voices in video games were often static and could not be updated to reflect changes to characters. They began working on a solution that would allow them to make dynamic, high-quality synthetic voices for characters and later expanded their offering to include other use cases.
The firm’s products are designed to optimize generated voices for virtual reality experiences, animated films and television, and audiobooks. Its software combines audio recordings with data on facial features and body movements to mimic an actor’s physical appearance, allowing the character’s voice to match its look. It is the same approach used by virtual assistants such as Alexa and Google Assistant to create their voice based on data about the person speaking.
But the potential for malicious actors to clone and manipulate voices is raising concerns. For example, in 2021, a documentarian sparked outrage when he used a company to clone the late celebrity chef Anthony Bourdain’s posthumous voice for a documentary. Ahmed says his firm’s products require explicit user consent to clone voices and adhere to strict usage guidelines to prevent malicious activity.
Other startups that nabbed recent funding include Seattle-area healthcare startup Aiberry, which is developing AI to detect mental health conditions by analyzing audio, visual, and language from a short, three- to five-minute conversation. It said the company will use the funds to expand its product into other markets and hire more employees. Other startups in the AI voice space recently include Hour One, Supertone, and Veritone.
