Weights.gg
Category: Voice / Speech AI
AI voice cloning social platform acquired by OpenAI in 2026, known for its consumer app Replay that allowed users to create and share synthetic voice models of celebrities and public figures. Weights.gg was founded in 2023. The company is led by JonLuca DeCaro. Based in San Francisco, California, United States. Team size: 1-10. Total funding raised: $4M. Latest round: Seed. Key investors include Original Capital; Precursor Ventures.
- Founded
- 2023
- Headquarters
- San Francisco, California, United States
- Team size
- 1-10
- Total funding
- $4M
Value proposition
Free, accessible AI voice cloning platform where users could create, share, and deploy AI voice models with high accuracy using consumer-grade hardware, operating as a social network for AI-generated voice content.
Products and solutions
Replay (consumer voice cloning app); Weights.gg platform (social network for creating and sharing AI voice models, text-to-speech, AI-generated images); AI voice model hosting and inference platform
Unique value
Democratized AI voice cloning through a social platform model — users could clone voices from short audio samples, share models publicly, and generate voice content for free, all optimized to run on consumer hardware via edge inference.
Target customer
Consumer creators, AI hobbyists, content creators, voice artists, and fans creating AI voice covers and synthetic voice content
Industries served
AI Voice / Speech; Generative AI; Social Media / Creator Economy; Entertainment
Technology advantage
Rewrote open-source AI models to run efficiently on consumer hardware (edge inference); multi-cloud infrastructure across AWS, GCP, and Azure with 9 clusters; 250+ concurrent GPUs; 10,000+ daily AI training jobs; 500,000 daily content creations; sophisticated spot instance optimization and GPU cost management
How they differentiate
Operated as a social network for AI voice models rather than a pure SaaS tool — users could browse, share, and remix voice models publicly. The platform hosted celebrity, character, and public figure voice models (Taylor Swift, Samuel L. Jackson, Donald Trump, etc.) in a community-driven marketplace format, differentiating from more restricted competitors like ElevenLabs.
Main competitors
ElevenLabs; Kits.AI; Uberduck.AI
Key partnerships
Northflank (infrastructure partner for multi-cloud GPU orchestration); AWS, GCP, Azure (cloud providers)
Notable customers
Late 2023: Founded as local AI voice cloning application; 2024: Launched web platform and Replay consumer app; Aug 2024: Raised ~$4M Seed round from Original Capital and Precursor Ventures; Scaled to millions of users with 500K+ daily content creations; Early 2026: Acquired by OpenAI (undisclosed terms); March 2026: Ceased operations; April 1, 2026: Platform officially shut down; May 2026: NYT reports acquisition publicly
Market positioning
Niche consumer AI voice cloning platform that gained popularity for hosting celebrity voice models, positioned between open-source voice tools and commercial services like ElevenLabs. Acquired by OpenAI primarily to remove legal liability around unauthorized voice cloning ahead of its planned IPO.
Geographic focus
United States (global consumer reach via web platform)
Patents and IP
Half a million content creations per day; 150TB monthly data transfer; Half a petabyte of user-generated content; 10,000+ daily AI training jobs; 250+ concurrent GPUs across 9 clusters
About JonLuca DeCaro
Ex-Citadel Engineer; Ex-Pinterest Engineer; University of Southern California (USC). Previously built and scaled complex systems handling millions of users at Citadel and Pinterest.