Clipto Raises $15M to Build Local AI Media Memory

Clipto disclosed $15M in all-equity funding at a $250M post-money valuation on August 31, 2026. The financing backs a local-first AI memory platform built to search video, audio, images, meetings, and documents across a user's own archive.

The investor group includes HSG, GL Ventures, EnvisionX Capital, Palm Drive Capital, Hans Tung, Lu Zhang, and 522 Ventures. Clipto says the capital will support on-device AI models, computing infrastructure that runs on consumer hardware, more integrations with AI agents, and global product expansion.

The market question reaches beyond one funding round. Generative AI has made it easier to produce media, while creators and knowledge workers still lose time finding the exact moment, quotation, decision, or source buried inside what they already made.

What Clipto Announced

The $15M financing disclosure arrived after a January announcement that Clipto had completed successive Pre-A, Pre-A+, and Pre-A++ financings since July 2025. That earlier company release said the latest funding moved Clipto's valuation above $250M but did not disclose an amount, making the August announcement the public source for the $15M figure.

That chronology matters because financing coverage can accidentally turn one capital sequence into several rounds. DevCuration is treating the August package as the verified $15M disclosure at a $250M post-money valuation and is not adding secondary database estimates or claiming a separate additional $15M close.

Clipto has not named a lead investor in its August release. The participating group spans HSG, GL Ventures, EnvisionX Capital, Palm Drive Capital, Hans Tung, Lu Zhang, and 522 Ventures, with Palm Drive and EnvisionX Capital independently confirming their investment relationships with the company.

From Stored Files to Usable Memory

Clipto's product begins with an ordinary failure of digital work: the file exists, but the person who needs it cannot remember the filename, folder, date, or application that owns it. A video editor may remember a shot, a lawyer a deposition answer, or a researcher a quotation without remembering the coordinates required to retrieve it.

The company indexes video, audio, images, meetings, conversations, and documents so users can search in natural language and return to exact source moments. Clipto's Model Context Protocol integration also allows connected AI tools to request source-linked results from folders that users approve, rather than receiving open-ended access to the entire computer.

Clipto describes its Mac and Windows products as local-first, with source media staying on the device by default. Its current website says mobile and web products are cloud-powered, so the more precise product promise is controlled choice over processing and access, not a universal claim that every workflow stays offline.

Why Henry Kang Kept Following Memory

Hongwen Henry Kang founded Clipto in 2023 with members of the founding team from his previous startup. Kang told TechCrunch that the underlying problem had followed him for roughly two decades, beginning with robotic systems at Carnegie Mellon that recorded their surroundings and remembered where objects were located.

Kang later applied related ideas to a startup that used AI to organize a user's closet and recommend outfits. His next company, Zenvideo, focused on making video creation easier and was acquired by Tencent in 2020, according to Kang's interview.

The progression gives Clipto a more coherent origin than a sudden pivot toward the latest AI category. The company is extending a long-running question about machine memory into a commercial layer for the media, documents, and conversations that accumulate around modern work.

Reported Traction Behind the Valuation

Clipto says more than 30M people have used its products and that it has hundreds of thousands of paying subscribers. Kang told TechCrunch that the company reached $15M in annual recurring revenue at the beginning of 2026 and remains profitable on a net-income basis, with just over 20 employees across the San Francisco Bay Area, Singapore, and Hong Kong.

Those metrics are company-reported and have not been independently audited in the reviewed sources. They still matter because a $250M post-money valuation against $15M in reported ARR frames this as a company financing an operating product, not a pre-revenue research promise.

Clipto initially focused on video creators managing footage across computers and external drives. Kang said creators now represent roughly one-quarter to one-third of users, while the rest include lawyers, doctors, researchers, marketers, HR professionals, professors, and students.

The Standalone Memory Bet

Adobe, Apple, and Google already provide natural-language search inside products and media libraries they control. Their structural advantage is distribution: search can arrive as another feature inside the editing, photo, storage, or operating-system workflow a customer already uses.

Clipto is making a different bet. It wants memory to cross file types, tools, and local archives, then make selected results usable by people and agents without turning the entire media collection into an unrestricted cloud data source.

The buyer is therefore not purchasing transcription alone. The buyer is paying to recover context, provenance, and usable source material from an archive whose storage problem was solved years ago but whose recall problem kept getting worse.

What the $15M Must Make Possible

Clipto says the financing will deepen the AI models and computing infrastructure required to run multimodal analysis on consumer hardware. It also plans to expand agent integrations and its international product footprint, with native Windows and Android applications joining Mac, iOS, and web access.

Execution will require balancing speed, accuracy, privacy, hardware limits, and product breadth. On-device models have less memory and compute than cloud systems, while customers will expect a search result to arrive quickly and point back to evidence they can trust.

If Clipto can make that tradeoff work across professional archives, the company can occupy a useful layer between storage and reasoning. The larger market will keep testing whether memory is durable independent software or a feature that platform owners eventually absorb, while every new recording, meeting, image, and generated clip keeps adding to the retrieval problem.