CaraOne finds your media. CaraOne XT puts it to work.
CaraOne is on-premises AI built for video: it watches, listens to and indexes your library so anyone can search it by meaning. CaraOne XT adds AskCaraOne, an AI agent that acts on that index to summarize, build stories and organize the workspace for you.
At a glance
CaraOne sees, hears and searches. XT adds an agent that reasons over what CaraOne found and acts on it.
| Capability | CaraOne | CaraOne XT |
|---|---|---|
| Understands visual content, speech and text; recognizes faces and voices | Yes | Yes |
| Indexes your library on-premises, from the folders you choose to watch | Yes | Yes |
| Search media by meaning, not just keywords | Yes | Yes |
| AskCaraOne Agent: plain-language requests that run multi-step tasks | No | Yes |
| Do Bar: filter, export and reorganize the workspace by request | No | Yes |
| Summaries, follow-up shot suggestions, proposed story structures | No | Yes |
| MAM integration and AI file and folder actions | No | Yes |
| Large language model (LLM) for reasoning | Not used | Local model, or your own OpenAI or Anthropic account |
| Audio, video or images sent outside your network | Never | Never |
How it works
Everything inside the dashed line runs on your systems. XT adds the agent; the external LLM is optional and receives text only.
Walkthrough: "Find interview clips where the subject discusses retirement"
The search itself always runs on your local index. The only thing that changes between setups is who interprets the request, and with an external LLM only the request text goes out.
| Step | CaraOne search | XT with a local LLM | XT with an external LLM |
|---|---|---|---|
| 1. Media in your watched folders is analyzed and indexed | On-premises | On-premises | On-premises |
| 2. Your request is interpreted | CaraOne's own models, on-premises | Local LLM, on-premises | Your OpenAI or Anthropic account |
| 3. Matching interview segments are found in the index | On-premises | On-premises | On-premises, guided by the LLM's text reply |
| 4. Results appear as thumbnails linked to the media | On-premises | On-premises | On-premises |
| What leaves your network | Nothing | Nothing | The text of your request |
Where XT earns its keep: story building
Ask AskCaraOne to "use these interviews to propose a story about retirement" and XT returns a suggested narrative structure built from your own footage.
Find
cNode, CaraOne's agent backend, finds the relevant interviews in your local index.Split
It splits the job into reasoning tasks: summaries, long-context analysis, narrative building.Reason
Those tasks run on your local LLM, or on your external LLM if you enabled one. With an external LLM, only selected text is sent: transcripts and scene descriptions.Return
cNode returns the proposed narrative structure to you inside CaraOne.The LLM only decides which CaraOne functions to call, based on text. It never receives audio, video or image files.
Your data, your control
External LLMs are optional, and nothing is sent to one unless you configure it.
Standard search never needs an external LLM
It runs on CaraOne's own models and your local index.Fully offline is available
XT can use local models for AskCaraOne, with no calls outside your network.Text only, when you allow it
With an external LLM enabled, only selected text is shared: your request, transcripts and scene descriptions. Audio, video and images are never sent to third-party providers.Your account, your terms
You connect your own OpenAI or Anthropic API account and pay the provider directly for usage. Data terms such as zero retention and no training are set in your agreement with that provider.One switch
External LLM use is turned on or off for the whole CaraOne system.Why not just use ChatGPT?
ChatGPT can't see your archive. CaraOne does the seeing, indexing, retrieval and workflow integration on your own systems, and XT uses a large language model only as a reasoning tool on top.
CaraOne uses internal models to understand and search media. AskCaraOne adds agentic workflows through our proprietary cNode backend, using local models or supported external LLMs based on customer configuration.
Deployment options
Pick the setup that matches your security rules. You can start fully offline and enable an external LLM later, or never.
| Setup | Where reasoning runs | What leaves your network |
|---|---|---|
| CaraOne | No LLM used | Nothing |
| CaraOne XT, fully offline | Local models on a dedicated server at your site | Nothing |
| CaraOne XT, hybrid | Your own OpenAI or Anthropic account | Selected text only |
Already running CaraOne?
Moving to XT keeps the work you have already done.
No re-indexing
Content you have already indexed carries forward to XT.XT is licensed per cluster
All nodes in a CaraOne cluster move to XT together. A separate cluster can stay on CaraOne, but the two will not share an index or search.Server sized for you
For fully offline XT, Scale Logic specifies the server for your workload and library.Talk to Scale Logic to plan your CaraOne or CaraOne XT deployment.
Book a demo and we'll walk through the setup that matches your security rules.
