Mimir introduces AI scene description to cloud asset management

Cutting speed and search accuracy are the fall colors for Mimir, with automated, detailed and accurate scene description, natural language search, advanced sharing, and quick cutting capabilities from archive to live.

Bergen, August 20th: Cloud-native asset management company Mimir announces all-new intelligent, fully automatic scene description capabilities to improve search and set speed-cutting records for its cloud video production and collaboration solution.

Context is everything

Finding relevant material in massive archives has long been a challenge that Mimir continues to innovate to address. By leveraging AI to accurately, automatically, and instantly detect and describe scenes in a video asset, Mimir now gives users rapid access to meaningful chunks of material from anywhere in their files and assets.

Leaning on semantic models to do the heavy lifting, the scene descriptions are automated, detailed, and accurate. This means an editor could search for a description of a scene that considers the full context of what is going on – say, “people waiting in line in an airport.” They are then presented with files with pre-chunked transcripts. These can easily be fully or partially dropped onto a timeline, edited, or exported, saving a lot of time.

Scene description in Mimir👇

A detailed view of Mimir scene descriptions

Semantic search for everyone

Beyond summarizing scene descriptions, semantic search capabilities allow users to search all their media assets and have results returned on the meaning of a query and not on its exact wording. A query can combine abstract concepts, visual context, and transcript-based content - all in a regular sentence in the searcher’s language. This natural language approach is the same one we are now accustomed to using when interacting with large language models.

Because Mimir customers have been used to automated AI-powered tagging for a long time, the shift for them will not be as tectonic as for media professionals used to less enhanced asset search. The difference will be in the user experience; the ability to use natural human language and improved quality of search results.

With semantic search, it’s unnecessary to be an archive aficionado, knowing exactly what to search for and what will get returned. For those editing a story, looking for archive, edited, near-live or live material to add to it, semantic search will open a world of content discovery. Journalists and editors can find any relevant material they did not know was in their archive and leverage the gold mine of historical content they can access. In Mimir itself, it will be possible to toggle between existing, term-matching tagged, and precise search and natural language semantic search, ensuring the right tool can be used in the right context.

Mimir leverages Amazon’s multimodal LLMs running on AWS Bedrock to power semantic search - but customers can also choose to integrate their preferred search providers. This flexible approach to integrating AI services combined with a continuous delivery method means Mimir can adapt to the rapidly changing reality of semantic search capabilities.

Semantic search in action👇

Quick cutting for fast delivery

Mimir has also further enhanced the quick-cutting capabilities introduced by the Mimir Cutter. Users can now add material from any source to a sequence to be rendered or delivered to an NLE, such as Adobe Premiere Pro, for full-feature editing.

By adding the capability to extract content from growing files as they are captured, Mimir makes it possible to mix and match any content in Mimir Cutter and Adobe Premiere Pro, from archive through edited, near-live, and live sources. This functionality adds value in both MAM, PAM, and DAM workflows, as the technical limitations sometimes found between different types of content are obliterated.

For users of Adobe Premiere Pro, Mimir will now work with partial downloads for high-res files - only downloading the sections of assets being worked on, thus increasing efficiency when working with larger assets by an order of magnitude.

More sharing, less resource waste

Taking collaborative working capabilities to the next level, Mimir users can now securely share, collaborate on, and receive any view, file, editing project, folder, or ongoing live ingest with other users in their organization or with external collaborators. Dynamic upload portals can be set up ad-hoc or permanently to allow stringers and freelancers to deliver time-critical material, one-off and in bulk, to the media organization.

A screenshot showing easy content request in Mimir

Collaboration on video projects across departments, easy handoff from one editor to another, and reuse and repurposing of assets and work are necessary for using creative resources to their fullest potential. Combined with fast-cutting capabilities and the paradigm-shifting addition of semantic search, Mimir makes it possible to create more, faster, and better.