OpenAI has reportedly acquired Weights.gg, a startup known for creating AI-powered voice cloning tools, a move that signals the company may be preparing for a much deeper push into synthetic voice technology.
The acquisition was not formally announced by OpenAI, but multiple reports indicate the deal happened earlier this year before Weights.gg shut down operations in March 2026. According to The New York Times reporting cited by Techmeme and LiveMint, OpenAI absorbed both the startup’s technology and its small team of employees, who have now reportedly been distributed across various internal OpenAI groups.
The terms of the acquisition have not been disclosed publicly.
Before shutting down, Weights.gg operated as a platform where users could create, share, and experiment with AI-generated voice models.
Its tools reportedly allowed users to generate convincing synthetic versions of celebrity voices, fictional characters, musicians, and political figures using relatively small audio samples. According to reports, the platform hosted AI voice models imitating public figures including Taylor Swift, Donald Trump, Joe Biden, and K-pop artists.
The startup also developed a consumer-facing app called Replay that simplified voice cloning workflows for creators and hobbyists. Rather than targeting enterprise clients directly, Weights.gg leaned heavily into internet creator culture and experimental AI media communities.
When users now visit the company’s site, they see a notice stating that Weights officially shut down on April 1, 2026.
The acquisition arrives just days after OpenAI launched a major new suite of realtime voice intelligence tools through its API platform.
Those new systems include:
OpenAI says the broader goal is to move voice AI beyond simple speech recognition toward systems that can “listen, reason, translate, transcribe, and take action as a conversation unfolds.”
The company increasingly appears focused on making voice a primary interface layer for AI products across customer support, assistants, live translation, events, education, enterprise software, and multimodal agents.
The Weights.gg acquisition likely strengthens that effort by adding expertise specifically around high-fidelity voice replication and synthetic speech generation.
What makes the acquisition particularly notable is that OpenAI has historically treated voice cloning as one of the riskiest categories in generative AI.
Back in 2024, the company publicly acknowledged it had developed a system called Voice Engine capable of generating highly convincing cloned voices from just 15 seconds of audio — but refused to release it broadly due to concerns around fraud, misinformation, impersonation, and election abuse.
At the time, OpenAI warned that synthetic voice tools could undermine trust in audio authentication systems and accelerate scams involving fake political speeches, fake emergency calls, and impersonation attacks.
That caution intensified after actress Scarlett Johansson accused OpenAI of using a ChatGPT voice that sounded similar to hers, leading the company to remove the “Sky” voice option from ChatGPT.
The company now appears to be walking a difficult line: advancing voice AI capabilities aggressively while trying to avoid becoming associated with uncontrolled deepfake ecosystems.
Voice cloning has rapidly evolved from a niche research area into one of the fastest-growing segments in artificial intelligence.
Companies like ElevenLabs, PlayAI, Resemble AI, and others have already built large businesses around synthetic voice generation for creators, gaming, media, dubbing, accessibility, and enterprise automation.
At the same time, regulators and copyright holders are becoming increasingly alarmed by how easily AI can now imitate public figures, artists, and copyrighted characters.
Several musicians and actors have already pursued legal protections around AI-generated replicas of their voices and likenesses.
That tension is turning voice AI into one of the most politically and commercially sensitive areas of the broader AI race.
Reports suggest OpenAI does not plan to relaunch Weights.gg as a standalone product. Instead, the technology and talent are expected to be folded into existing internal voice initiatives.
That could eventually influence everything from ChatGPT voice assistants to realtime translation, AI companions, enterprise voice agents, creator tools, and multimodal AI systems.
The broader industry trend is becoming increasingly clear: AI companies no longer see voice as just an accessibility feature.
They increasingly view it as one of the most natural interfaces between humans and intelligent systems.
And as those systems become more realistic, expressive, and personalized, the companies controlling voice generation technology may end up shaping how people interact with AI itself.
Mirami Chat showed up in my research feed three different times last m...
If you've ever searched for today's Wordle answer, a crossword clue yo...
Try this on someone tonight. Lean in slightly, raise an eyebrow, and a...
For years, the biggest fear surrounding artificial intelligence was th...
A quick experiment. Open itsnewztalkies.com in one tab. Open the third...
A new AI startup founded by former Google DeepMind researchers is rapi...
Discussion