Many AI tools that can be seen on this day and summarize what they will do, but it is a bit difficult when you get questions about multiple videos and snapshots.
This is a great limit for security companies you want to use AI to scrub from different cameras from different cameras, as well as a marketing company you want to learn video campaigns and peak products.
Memories Want to solve the problem with ai platform that can process a 10 million hours video. For companies with many videos to analyzing, beginners want to provide contextual layers, complete with indexing, tags, segment and aggregations.
Pure folors Dr Wahwah Scientists in the Meta of Meta when she pursued PhD, and the enmantrae – Ben) partnerships in Meta is the engineering learning machine.
“All the top AI companies, such as Google, Opening and Meta, have been focused to produce the last model. The Shen,” Shen Shen, “said Shen Techcrrunch.
“But when humans use the visual memory, we complete the great context. We are inspired by this and want to build a solution to know the video around the world,” he said.
To the purpose, the company has now increased $ 8 million in the seed led by Besar, and with the participation of the Samsung Next, Crane, Crane, and creator. Shen said the company with the purpose of increasing $ 4 million, but it was done with a circle of uselessness due to investors.
TechCrunch Events
San Francisco
|
October 27, 2025
“Shen is a highly high technical founder, and she wants to see the boundaries of video and intelligence,” said Misha Gordon-Rowe, his friend in the Acentures in Susa. “Memories.ai can unlock the first visual intellijen data with a gap in the market for the intelligence of the context, which attracts to invest in the company,” he said.
Samsung Next has a slightly different thesis: Samsung’s investment arm looks for memories.As useful for consumers.
“One thing you like about the memories of memories is to do a variety of devices computing. This means you don’t have to save better security data to make better security data,” SAM Campbell, partner on Samsung Next.
Memories.ai said using his own pile of technology and model for analyzing. First, remove the sound of the video and through the output through the coating compression only store important data. Then there is a layer of indexing, which makes video data searched (using a natural language question) with segments and tags. There are also aggregate layers that summarizes the data from the index, helping to make a report.
Currently, Introduction Caters for two types of companies: marketing and security. The marketing companies can use the starting tool to find trends associated with brands on social media, and identify social videos, and identify the video you want to do. Memories.ai also provides a tool for marketers to make the video.
This company can also be used with security companies to help analyze security trap to determine the action that can be dangerous by people in the video with considerations through the pattern.
Now, the companies work with the.ai companies must upload the video library to the platform to help analyze the clip. But Shen said to the forefront, the clients will be able to make the drive drive and syncing the content as much as possible. The plan is to enable customers asking as: “Tell me everything about someone who is interviewed in the last week.”
Envisions Shen are AI assistant that can earn context of user user via photos or when enabled smart enabled sunglasses. They also see the technologically played a role to train the humanoid robot to perform complex tasks or helps the driving cars of the driving independently reminds a different route.
Now the company now has 15 people, and plans to use funds to add a team and add a search.
Memories.ai are the same startup, as MeM0 and ReadIt is used to provide memory layers for ai model, although it provides limited video support at the moment. It also has to quarrel with companies like Relazel and Google, which has been used to help model AI knowing the video.
However, as well as the company is more horizontal, which can also be used with different video models.