Synology Drive AI Search: Find Files by Description

Have you ever known exactly what a file contained but had no idea what it was called?

Maybe you remember a photo showing someone standing beside a red truck, a PDF discussing next year’s storage budget, or a recording in which someone talked about a particular customer. Traditional NAS search often forces you to remember filenames, folders, or exact keywords before you can find any of those files.

Introduced as part of the DSM 7.4 generation, AI Search allows users on supported GPU-equipped Synology systems to search Synology Drive using natural-language descriptions. Instead of matching only literal words, semantic indexing can analyze the meaning and content of documents, images, audio, and video.

For organizations with large collections of poorly named or unstructured files, that could make finding information substantially easier.

What Is Synology Drive AI Search?

Traditional search works best when you already know something concrete about the file.

You might search for:

Q4-budget.xlsx

or:

network upgrade

AI Search lets you approach the problem differently.

You could instead enter something closer to:

“Presentation discussing next year’s network expansion budget”

Synology Drive then uses semantic meaning to rank files that may be relevant, even when your exact search phrase does not appear in the filename or document.

Synology calls this semantic indexing.

Once enabled, Synology Drive also provides an AI Highlights area in search results, allowing semantically relevant files to surface separately from conventional keyword matches.

Find Documents Without Remembering Their Exact Words

Document search is one of the clearest practical uses.

Imagine an employee remembers reading a PDF about replacing aging office servers but cannot remember:

The filename
Who created it
Which folder contains it
The precise wording inside it

Instead of trying dozens of keyword combinations, the employee might search:

“Proposal about replacing old servers and reducing maintenance costs”

The embedding model analyzes document context and intent rather than relying completely on direct keyword matches.

Synology’s semantic indexing currently supports common document formats including Microsoft Office files, Synology Office documents, PDFs, and plain-text formats.

That makes AI Search particularly useful for businesses with years of project documents, reports, proposals, technical documentation, and archived files.

Search Images by Describing What Is in Them

The capability becomes even more interesting with images.

Synology’s AI Runtime includes an image captioning model that generates descriptive information about images and video frames.

That can make visual content searchable based on what appears in the image rather than merely its filename.

For example, instead of trying to remember whether a photo was called IMG_3948.jpg, a user could search for something descriptive such as:

“Photo of equipment racks in a server room”

or:

“Picture of a whiteboard with network diagrams”

That is much closer to the way people actually remember photographs.

For businesses managing inspection photos, project documentation, marketing assets, site surveys, or equipment images, descriptive search could eliminate a considerable amount of folder browsing.

OCR Makes Text Inside Images Searchable

Image recognition is only part of the system.

Synology also provides an OCR model, or Optical Character Recognition.

OCR extracts text found inside:

Images
Scanned documents
Video frames

If OCR recognizes information within the scan, however, users have a much better chance of finding it based on the document’s actual content. This could be useful for scanned contracts, receipts, invoices, handwritten-to-digital workflows, screenshots, archived paperwork, and other image-heavy business records.

Audio and Video Can Become Searchable Too

Synology’s on-device AI stack also includes speech-to-text, or STT. This model converts spoken words in supported audio and video files into text that can contribute to search results.

Imagine searching a collection of meeting recordings for:

“Discussion about migrating VMware servers”

Rather than opening every video individually, indexed speech can help Synology Drive identify recordings containing relevant discussion. Semantic indexing currently supports selected audio formats including MP3, WAV, and M4A, along with video formats such as MP4, MOV, MKV, and WebM.

This could make large archives of meetings, interviews, training materials, and recorded presentations substantially easier to navigate.

AI Search Runs Locally on Supported Synology Hardware

An important part of Synology’s approach is that the AI models powering this feature can run on the NAS itself.

Synology AI Runtime provides local inference for:

Embeddings
OCR
Speech-to-text
Image captioning

That means the core semantic processing does not inherently require sending the files being indexed to a public AI service.

For businesses dealing with confidential documents, internal recordings, customer information, or proprietary images, keeping AI processing close to the stored data can be an important advantage.

You Need a Supported GPU-Enabled Synology System

This is the most important limitation.

Synology requires supported storage systems with compatible GPU hardware for its on-device AI models. The current AI Runtime beta applies only to a limited set of higher-end systems, including supported FS and RS models, and requires a compatible GPU.

Synology also recommends allowing substantial system memory for AI workloads. Its technical specifications estimate approximately:

2 GB for the embedding model
2 GB for OCR
3 GB for speech-to-text
3 GB for image captioning

Because DSM and other applications also require memory, Synology recommends reserving at least 16 GB of system memory when using these models. So this feature should not be viewed as something every two-bay DiskStation can simply enable.

Administrators Must Enable Semantic Indexing

AI Search also requires configuration. After installing the appropriate GPU and AI packages, administrators enable the required models within Synology AI Console.

They then configure indexing through:

Universal Search → Preferences → File Indexing

For the desired Synology Drive Team Folder, the indexing method must be changed to Semantic indexing.

This is important because semantic indexing is currently configured specifically for Synology Drive Team Folders. Administrators should therefore decide which folders actually benefit from AI processing rather than indexing everything simply because they can.

There Are File Size and Storage Limitations

Semantic indexing has practical limits.

Current Synology specifications list maximum semantic-indexing sizes including:

Office documents and PDFs: 30 MB
Plain text and Markdown: 200 KB
Images: 20 MB
Audio: 100 MB
Video: 200 MB

If a supported file exceeds the semantic indexing limit, DSM can fall back to conventional file-content indexing instead. Semantic indexing is also not currently supported for Hybrid Share folders, remote folders, cloud-mounted storage, or external USB/SD storage. These restrictions matter when deciding which datasets should be included.

AI Search Is Different From Universal Search

There is another easy point of confusion.

Synology has offered Universal Search for years, and Universal Search continues to provide filename, metadata, and full-text indexing across DSM. It adds semantic understanding to Synology Drive, allowing people to search according to meaning rather than exact terms.

Synology’s current specifications specifically state that AI Search is supported in Synology Drive rather than as a general AI-search replacement for Universal Search throughout DSM.

Is Synology Drive AI Search Worth Using?

For the right environment, yes.

It is particularly compelling when an organization has thousands or millions of files whose names and folder structures no longer make finding information easy. The biggest advantage isn’t simply “AI.” It is the ability to search closer to the way humans remember information.

You may forget IMG_00457.jpg but remember “photo of the damaged server cabinet.”

You may forget meeting-final-v3.mp4, but remember “meeting where we discussed replacing VMware.”

And you may forget which PDF contained a recommendation, while still remembering what the recommendation was about. That is where semantic search can make Synology Drive genuinely more useful.

As of September 2026, however, businesses should remember that Synology Drive 4.1 AI Search and its supporting on-device AI components remain part of Synology’s staged/beta rollout for compatible GPU-equipped systems. Organizations should verify hardware compatibility and production-readiness before deploying beta packages on business-critical storage. Synology consulting services can help evaluate compatible hardware and AI workloads.

About Epis Technology

Epis Technology helps businesses design, deploy, upgrade, and optimize Synology environments around real storage and workflow requirements. Services include Synology NAS consulting and support, DSM configuration, Synology Drive deployments, storage architecture, GPU-capable systems, enterprise backup, virtualization, Microsoft 365 and Google Workspace protection, networking, ransomware resilience, and disaster recovery planning. Epis Technology can help organizations determine whether Synology Drive AI Search fits their environment and correctly plan the hardware, memory, indexing, permissions, and data-protection strategy required to deploy it.