The following are very general questions and answers gathered from users.  Contact us if you don't see your question here.  There is no general ordering to this list.

Is CaraOne only available via the Scale Logic AI Nodes or will there be reference Architecture? 

  • Customers must buy a complete appliance loaded with CaraOne and ready for deployment. This is due to the specific CPU and GPU workload and a supported worldwide SKU.  Cloud-only installs can be discussed with our team for large deployments.

What happens to a scan if the node crashes or has a power outage halfway through?

  • It just continues as it is stateless.  It keeps track of its progress and starts where it left off.  The DB is stateless as well.

Can multiple AI nodes be set up for high availability and scaling?

  • Yes.  Scaling needs further explanation.  Adding servers for scaling is done in practice in one of 2 ways.  First, it can reduce the watch and analysis time.  2x the servers make it half. 10x servers 1/10th, 20x servers 1/20th.  A single server can watch content 3 to 14 times in real-time.  You can use this figure to estimate how fast your library could be available for searching with these numbers.  The second is additional servers can be dedicated to searching and the predictive timeline tool.  Searching is set at 20 users per server, however, the predictive timeline requires more horsepower.  We can set up server(s) dedicated to tasks to keep new analyzed content available as fast as you need it and ample performance for searching and the predictive timeline.  Additionally, analyzing can be setup at night so that servers can be available for search and predictive timeline during the day.

Can remote staff access the CaraOne server?

  • Yes via a VPN.  A company's existing VPN puts a remote CaraOne client on the same network as the CaraOne server.  The proxy files are 4-5 Mbit.

As CaraOne already knows very famous people but can't know most people, can you teach CaraOne additional facial recognition?

  • Yes via a folder called "faces".  There you can create a folder for each name spelled exactly how you want to search, and place 5 to 6 portrait shots of various variety.  These photos can only have one face per picture so you will need to crop in some cases.  There should not be blurry portraits, sunglasses, or face coverings.  More than x500 resolution portraits have been shown to get the best results.  CaraOne can even understand faces with varying ages, as the person has aged.  When using nicknames or similar names they must be unique.  CaraOne does not have to re-index to put faces to the name.   

Can CaraOne be directed to proxies already created?

  • Yes.  There are times when there is no reason to recreate proxies and it makes sense to point to the existing proxies.  If you don't have proxies already, CaraOne automatically creates H264 for more efficient browsing and playback.  Whenever clips are found and the "action" button is pressed, CaraOne always points to the high-resolution version.

My storage is not Scale Logic or is cloud based, can CaraOne watch it?

  • CaraOne can analyze any on-prem or cloud storage that can be mounted on a Linux OS as a file system.  This enables a huge variety of storage technology.  Check our installation KB articles for specific storage setup processes.  We recommend Avid Nexis or SMB NAS storage however even SAN storage can work.  Cloud buckets need to be mounted as a file system on the CaraOne Linux servers.   Archive storage likely will need to be the HSM variety that historically mounts as a mount point or drive letter.  Archives that can't be mounted as a file system will need a "rehydrate" method applied.  Check with Scale Logic on this process as we have experience monetizing archives.

What search engine are you using under the hood to search content from metadata?

  • CaraOne uses its own search engine that is tuned to the neural network supporting CaraOne's incredibly fast search.   As CaraOne watches the content, it's more about understanding the meaning, expression, and emotion of the video.  The search engine or the CaraOne search API has to be used.

Which server APIs (or models running locally) are you leveraging to extract machine learning (ML) features from video, faces, text files, audio, and spoken word?

  • Our own. There are no external APIs used.  When the inventor set out 6 years ago with the data scientists, they had to design all their own tools as they didn't exist yet.  They are all installed on the server when you receive it, and updated via software updates we offer as download links on supported systems via support contracts.

What are the patching and update processes in terms of access to a customer’s on-prem system? 

  • CaraOne can be installed completely air-gapped.  We will provide updates via file downloads that can be copied to the CaraOne server for updating.  CaraOne can NOT access the internet to "learn" new things.  This is a completely on-prem solution dedicated to the isolation of your media.

Is there any charge for other languages?  How many languages do you currently support?

  • We support 150+ languages. However, there is (at the moment) just one language per system.  We endeavor to enhance this via the road map. 

Can you cater your searches to specific demographics? (Male/Female/Age?)

  • Yes

What NLE editors do you support? Adobe, Avid Resolve. Others?

  • Currently CaraOne interfaces with Premiere and Avid Media Composer.  Resolve and FCPX can be supported via AAF import.  With Premiere we interface with the already installed Premiere REST API.  With Avid we analyze the projects and bins to learn their contents.  With Resolve and FCPX we can't read inside (yet) but you can export an AAF and import it into Resolve.  With FCPX you have to further convert the AAF into XML for import.

What is the speed of the CaraOne Server? Is there a point of diminishing returns? 

  • Based on an appliance with a single GPU, the appliance is capable of doing between three to fourteen times in real-time. For example, if you have 8 hours of content being ingested it will be completed in 3 to less than an hour depending on the content. You can scale the appliances linearly to increase performance.  Another way to think of this is that 100 hours of video could be completed in as little as 14 hours per server.  The servers scale linearly.

How does the annual support work?

  • The first year is included in the purchase of the appliance.  You can add up to 3 years to the initial quote or add support year by year.  The cost of annual support from Scale Logic is always affordable.  One contract covers HW break/fix, CaraOne software assistance, and CaraOne software upgrades including the models and new roadmap features upon release.

Can CaraOne watch growing files or files that are still being recorded?

  • No.  The file will have to be closed out on the file system first.  CaraOne's built in fswatcher will see it and start to analyze it a few minutes after the record is finished.