Many AI devices can watch a video today and briefly tell what is going on, but when you ask questions of models spread for many hours about many videos and footage, things become a bit difficult.
This is a large limit for security companies that want to use AI through thousands of hours of footage from various cameras, as well as marketing companies who want to study various video campaigns and product shoots.
Memories. Wants to deal with the problem with its AI platform that can process videos up to 10 million hours. For companies with lots of videos to analyze, the startup wants to provide a relevant layer, which is completed with searchable sequencing, tagging, segment and aggregation.
Its co-founder Dr. Sean Shen was a research scientist in the reality labs of Meta, when he was chasing his PhD, and his counterpart enmines (Ben) Zhou worked as a machine learning engineer in Meta.
“All top AI companies, such as Google, Openai and Meta, are focused on producing end-to-end models. They are good, but these models often have limitations around understanding video reference beyond one or two hours,” Shane told Techcrunch.
“But when humans use visual memory, we squeeze through a large reference of data. We were inspired by this and wanted to create a solution to better understand the video in several hours,” he said.
On that target, the company has now raised $ 8 million in a seed funding round led by Susa Ventures, and with the Samsung Next, Fusion Fund, Crane Ventures, Seedcamp and Participation from Manufacturer Ventures. Shane said that the company had initially targeted to raise $ 4 million, but the investor ended with an oversbird round due to interest.
Techcrunch event
San francisco
,
27-29 October, 2025
“Shane is a high-technical founder, and he is obsessed to carry on the boundaries of video understanding and intelligence,” said Misha Gordon-Rowe, a partner of Susa Ventures. He said, “Memories. A lot of first-a-sight visuals with their solutions can unlock intelligence data. We felt that there was a difference in the market for longer visual visual intelligence, which attracted us to invest in the company,” he said.
Samsung Next had a slightly different thesis: Samsung’s investment hands over the hands of the hand. Solve to be useful for consumers.
“One thing we like about memories. AI is that it can do too much on-device computing. This means that you do not need to store video data in the cloud. It can unlock better security applications for those who are apprehensive by installing security cameras in your home due to privacy concerns,” said a partner of Samsung.
Memories. First, it removes the noise from the video and passes the output through a compression layer only to store vital data. Then there is a sequencing layer, which creates video data with partitions and tags to find (using natural language queries). There is also an aggregation layer that summarizes data from the index, helps in making reports.
Currently, the startup completes two types of companies: marketing and safety. Marketing companies can use startup equipment to see the trends related to their brands on social media, and recognize what kind of videos they want to make. Memories. AI also provides equipment for abusive to create those videos.
The company is also working with security companies to help them analyze safety footage to determine potentially dangerous tasks by people in the video by arguing through patterns.
Currently, companies working with memories need to upload their video library on the platform to analyze the clip. But Shane said that in the future, his customers will be able to create a shared drive and sync the material more easily. This plan is to enable customers to ask questions: “Tell me about all the people I interviewed in the last week.”
Shane applies an AI assistant that can get references through his photos on the user’s life or when they activate smart glasses. He also sees that technology helps to remember different routes or remember self-driving cars, playing a role in training Humanoid robots to perform complex tasks.
The company currently employs 15 people, and is planning to use funds to increase its team and improve its discovery.
Memories. mem0 And Lata, which are working on providing a memory layer for the AI model, although they currently provide limited video support. It will also have to struggle with companies such as Twelvelwell and Google, who are working on helping AI models understand the video.
However, Shane feels that his company’s solution is more horizontal, which will also allow it to work with various video models.
