Skip to content

Serverconfig.xml

ServerConfig.xml holds all the server's settings. It is stored in C:\ProgramData\FastVideoCataloger\ServerConfig.xml and is read once, when the server starts.

In most cases you should not edit this file by hand:

  • The Share Video Catalog wizard (Start → Server → Setup) writes it for you.
  • Admin Server (Start → Server → Admin Server) edits the same file through a user interface. Saving there stops the service, writes the file and starts the service again, so your changes take effect immediately.

If you do edit the file by hand, restart the Fast Video Cataloger service afterwards, or the server will keep running with the old settings. If the console server is started without a serverconfig.xml file, it will create one for you after you answer some questions.

Catalog

Setting Default Description
<CatalogFile> (none) Fully qualified local path to the .vcat file to host. The server reads and writes exactly the same file format as the standalone program, so a catalog built on a single machine can be shared as it is. A server hosts one catalog — see Working with more than one catalog. The file must be local to the server, see Distributed file system.
<CatalogPassword> (empty) Password for a password protected catalog. A password protected catalog cannot be shared — remove the password from the Info menu's Set password page, by entering an empty password, before you share it.
<LicenseKey> (empty) The Fast Video Cataloger license key this server activates against. It decides how many catalog user accounts you may create. An empty key still serves the catalog, it only caps new accounts.

Network and file hosting

Setting Default Description
<NetworkPort> 8754 The port the server opens for clients. Clients connect to machine:port, for example YourComputer:8754. All communication except large file transfers goes over this port.
<AllowRemoteConnections> false When true, the server accepts connections from other computers. When false, only connections from the same machine are accepted. The sharing wizard sets this to true.
<EnableBuiltInFileHosting> false When true, the server serves videos, companion images and actor photos over HTTP itself, with no external web server needed. The wizard sets this when you pick built-in hosting.
<RequireAuthForFileHosting> true When true, a client needs a catalog account to fetch files from the built-in HTTP server.
<UseIdBasedStreaming> false When true, clients stream videos through /api/v1/videos/{id}/stream instead of by file path. The server looks the path up in the catalog, so video files do not need a common root folder and can live anywhere the server can reach. The wizard sets this together with built-in file hosting.
<NetworkDriveMappings> empty Written by the Share Catalog wizard. One <NetworkDriveMapping> per network drive letter the catalog's video paths use, with <Drive> (S:) and <UncRoot> (\\nas\videos). A mapped drive letter only exists for the user who mapped it, not for the service. With <UseIdBasedStreaming> on, the server hands clients the UNC path of a video on such a drive, or of a video stored with a UNC path, and the desktop client plays it from the share when it can reach it and streams otherwise. The stream endpoint also tries the UNC path, which works when the service logs on as an account with access to the share.

Webhooks

Setting Default Description
<WebhookUrls> (empty) Addresses the server sends video events to, separated by spaces, commas or semicolons, for example https://hooks.example.com/fvc. Empty sends nothing. See Webhooks for the events. Restart the server after changing it.
<WebhookSecret> (empty) When set, each webhook request is signed with it in an X-FVC-Signature header, so the receiver can check that the call came from your server.

Proxies

A proxy is a smaller H.264 copy in MP4 that the server makes of a video a browser cannot play, such as WMV, AVI, ProRes, MXF or DV, so the web client can play it. Proxies are made by a program the server runs once per video.

Setting Default Description
<ProxyFolder> C:\ProgramData\FastVideoCataloger\Proxies Where proxies are kept, in a subfolder named after the catalog, one <video id>.mp4 per video. Empty turns proxies off. The server only ever deletes files in this folder.
<ProxyMaxSizeGB> 20 The largest the catalog's proxy folder may get. When a new proxy would pass it, the least recently played proxies are deleted first. 0 is no limit. A proxy takes just under 1 GB per hour of video with the default arguments.
<ProxyAfterIndexing> false When true, a video that needs a proxy gets one as soon as it has been indexed. Otherwise a proxy is made the first time the video is played in the web client, or when asked for through the REST API.
<ProxyProgram> (empty) The program that makes a proxy. Empty uses the mpv.exe installed with the server.
<ProxyArguments> (empty) The program's arguments. Empty uses the default arguments below, also when a later version improves them. {video_file} becomes the video's path and {proxy_file} the file to write, both already in quotes. A proxy counts as made when the program ends with exit code 0 and has written the file.

The default arguments make a 720p proxy at 2 Mbit/s with the H.264 encoder built into Windows, at the video's own frame rate up to 60, or 30 when the file does not state one:

--no-config {video_file} --o={proxy_file} --of=mp4 --ofopts=movflags=+faststart --ovc=h264_mf --ovcopts=b=2M --oac=aac --oacopts=b=128k --vf=scale=-2:720,lavfi=[fps=fps=if(between(source_fps\,1\,60)\,source_fps\,if(gt(source_fps\,60)\,60\,30))]

Windows Server and the N editions of Windows do not have that encoder until the Media Foundation feature (Server) or the Media Feature Pack (N editions) is installed. The reason a proxy failed is in ProxyService.log in the server's program folder and in the proxy's error in the REST API.

Video, photo and actor photo locations

These settings map a file's location on the server to the path a client uses to read it. They are only needed when clients access files by path, that is over a Windows share or an external web server. With <UseIdBasedStreaming> turned on they are optional.

Setting Description
<VideoFileLocalRoot> Local path on the server to the folder that is the root of all video files.
<VideoFileShareRoot> The path clients use to reach that same folder, for example \\fredrik_asus\videocatalog\videos for a Windows share, or an http:// address for a web server.
<PhotoFileLocalRoot> Local path to the root of all video companion images.
<PhotoFileShareRoot> The path clients use to reach the companion images.
<ActorPhotoFileLocalRoot> Local path to the root of all actor companion images.
<ActorPhotoFileShareRoot> The path clients use to reach the actor images.
<VideoFileLocalRootNoCloud> <PhotoFileLocalRootNoCloud> <ActorPhotoFileLocalRootNoCloud> The same local roots with any cloud drive (Dropbox, OneDrive) path resolved to a real folder on disk. The wizard fills these in when a root is on a cloud drive, and the server uses them in preference to the plain roots when serving files. Leave them empty otherwise.

If a root is missing or unreachable when the server starts, the server writes a warning to its log and keeps running. See Troubleshooting for what that means for the files under it.

Uploads

Setting Description
<ServerUploadFolder> Local folder that receives uploads from clients while the transfer is in progress. Failed uploads are left here, so with the server stopped this folder should be empty.
<VideoUploadDestination> Folder that finished video uploads are moved to. This must be a real folder, not a drive root. If it is empty, uploaded videos stay in <ServerUploadFolder>.

Indexing

The server can index videos added to it. These settings match the equivalent preferences in the desktop application.

Setting Default Description
<ThumbnailWidth> 320 Width in pixels of captured thumbnails.
<SampleStart> 60 Seconds into the video of the first thumbnail.
<SampleTime> 60 Seconds between captured thumbnails.
<MinThumbnails> 2 Fewest thumbnails per video; a video too short for that many is sampled evenly instead.
<MaxThumbnails> 0 Most thumbnails per video. 0 means no maximum.
<DvClipThumbnails> true For a DV camcorder capture, take one thumbnail at the start of each recorded clip, found from the camera's recording time, instead of sampling on an interval. At most <MaxThumbnails> clips, evenly spread.
<UseDirectshowMerit> false Use the decoders registered on the system instead of the bundled ones.
<LAVSplitterPath> LAVFilters\LAVSplitter.ax Splitter used for decoding. A relative path is resolved from the Fast Video Cataloger program folder.
<LAVDecoderPath> LAVFilters\LAVVideo.ax Video decoder used for decoding.
<ExtractMetaNfo> true Read the title, plot, link and tags of the .nfo file beside the video. The title, plot and link are set when the video is indexed for the first time, so a re-index keeps a title typed since; the tags are added every time.
<ForceFilenameTitle> true Keep the file name as the title: a title in the video file's own metadata (shell or XMP) does not replace it. A title in the .nfo still applies.
<AutoRotateThumbnails> true Store the thumbnails of a video recorded turned, such as a phone video, upright, by the rotation in its metadata. Needs <ExtractMetaMediaInfo>.
<ExtractMetaShellInfo> true Read Windows shell metadata from the video file.
<ExtractMetaXMPInfo> false Read XMP metadata from the video file.
<ExtractMetaMediaInfo> true Read media metadata such as resolution, codec and frame rate.
<ExtractSRT> true Import the .srt file beside the video as its transcript. A drone's flight log in an .srt is skipped, a transcript corrected by hand is kept, and with transcription on, a transcription that finds speech replaces it.
<AddTagsFromPath> true Add keywords from the words of the video's folder path and file name.
<AutoTagAllowNew> false Let a path word that is not a keyword in the catalog yet become one. Off, only the catalog's keywords are added.
<AutoTagMinTagLength> 3 Fewest characters a path word needs to be a keyword.
<AutoTagAllowNumbers> false Let a path word that is a number, such as a year, be a keyword.
<AutoTagTrimStartNumbers> true Remove the numbers at the start of a path word, so 2021Summer is Summer.
<AutoTagTrimEndNumbers> true Remove the numbers at the end of a path word, so Summer2021 is Summer.
<AutoTagInBrackets> false Take only the words inside (...), [...] and {...} in the path.
<IgnoreKeywords> common words Path words never taken as keywords, one <string> element each. Left out, the server ignores the same common words as the desktop (the, and, of...). The Share Catalog wizard copies the desktop's list.
<RemoveSimilarThumbnails> false Discard thumbnails that are near duplicates of the previous one.
<RemoveSimilarThreshold> 0.90 How alike two thumbnails must be to count as duplicates.
<RemoveFacelessThumbnails> false Discard thumbnails that contain no face.
<RemoveFacelessSkipVideosWithoutFaces> true Leave a video alone when none of its thumbnails has a face, instead of removing all but one.

AI during indexing

All AI features are off by default and need their models downloaded before they can be enabled. Model paths that are not absolute are resolved from the Fast Video Cataloger program folder. If a model is missing, the server logs it, skips that step and carries on indexing.

Setting Default Description
<EnableFaceRecognition> false Detect and recognise actor faces while indexing.
<FaceRecognitionConfidenceThreshold> 0.64 How closely a person in the video, averaged over their faces, must match an actor to be tagged. Before 10.5 the default was 0.62; a configuration still holding it is moved to 0.64 once.
<FaceRecognitionMinConfidence> 0.60 The lowest match a single face needs to count toward an actor. Used only when the YuNet and SFace models are missing and single faces are tallied; not shown in Server Setup.
<FaceRecognitionMinFrames> 2 Faces a person in the video must be seen in before they can be tagged as an actor.
<FaceRecognitionRuleVersion> 0 Set by the server, not a setting: 2 once the face settings have been moved to 10.5's rule.
<FaceRecognitionFrameSkip> 1 Process every n:th frame. Raise it to trade detail for speed.
<FaceRecognitionModelPath> FaceModels Folder holding the face models.
<FaceRecognitionUseCnn> false Use the slower, more accurate face detector.
<EnableSceneClassification> false Classify what kind of scene a thumbnail shows.
<EnableObjectDetection> false Detect objects in thumbnails.
<SceneClassificationMinConfidence> 0.70 Lowest confidence that still produces a tag.
<SceneClassificationFrameSkip> 1 Process every n:th frame. Raise it to trade detail for speed.
<SceneClassificationModelPath> SceneModels Folder holding the scene and object models.
<ObjectDetectionModel> yolox_l.onnx File name of the object detection model inside that folder. A custom imported model can be named here instead.
<EnableImageTagging> false Tag thumbnails with the image tagging vocabulary while indexing.
<ImageTaggingSensitivity> 3.0 Higher values give fewer, more confident tags.
<ImageTaggingModelPath> TrainedTagModels Folder holding the image tagging preset and the SigLIP encoder it shares with trained tags. TagModels, the default up to 10.3, is still used while it holds a preset and the shared folder does not.
<TagPacks> (empty) Extra vocabulary packs to enable, comma separated. The core vocabulary is always active.
<EnableTrainedTags> false Match thumbnails against the catalog's own trained tags while indexing.
<TrainedTagModelPath> TrainedTagModels Folder holding the SigLIP encoder shared by trained tags, image tagging and the visual search index.
<EnableThumbnailEmbeddings> true Store a compact description of every thumbnail while indexing, for visual search. Needs the encoder in the folder above; silently off without it.
<EnableTextRecognition> false Read the text on screen in every thumbnail's frame while indexing, with the Windows built-in OCR, so scenes can be searched by it.
<TextRecognitionLanguage> (empty) OCR language tag, such as en-US. Empty uses the service account's Windows language. The language's OCR must be installed in Windows on the server; without it no text is read and the server log says so.
<EnableTranscription> false Transcribe speech in videos while indexing.
<TranscriptionModelPath> WhisperModels\ggml-base.en.bin The speech to text model file to use.
<TranscriptionLanguage> auto Spoken language, or auto to detect it per video.
<TranscriptionTranslateToEnglish> false Write the transcript in English whatever language is spoken. Needs a multilingual model.
<TranscriptionDetectSpeakers> true Tell the voices in a transcript apart, so each line knows who speaks it.

Example

A minimal configuration using the built-in HTTP server:

<?xml version="1.0"?>
<ServerConfiguration xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns:xsd="http://www.w3.org/2001/XMLSchema">
  <CatalogFile>c:\videocatalog\test.vcat</CatalogFile>
  <NetworkPort>8754</NetworkPort>
  <AllowRemoteConnections>true</AllowRemoteConnections>
  <EnableBuiltInFileHosting>true</EnableBuiltInFileHosting>
  <RequireAuthForFileHosting>true</RequireAuthForFileHosting>
  <UseIdBasedStreaming>true</UseIdBasedStreaming>
  <ServerUploadFolder>c:\videocatalog\uploads</ServerUploadFolder>
  <VideoUploadDestination>c:\videocatalog\videouploads</VideoUploadDestination>
</ServerConfiguration>

The same catalog served over a Windows share instead, where every video file lives under one common root folder:

<?xml version="1.0"?>
<ServerConfiguration xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns:xsd="http://www.w3.org/2001/XMLSchema">
  <CatalogFile>c:\videocatalog\test.vcat</CatalogFile>
  <VideoFileLocalRoot>c:\videocatalog\videos\</VideoFileLocalRoot>
  <VideoFileShareRoot>\\fredrik_asus\videocatalog\videos</VideoFileShareRoot>
  <PhotoFileLocalRoot>c:\videocatalog\photos\</PhotoFileLocalRoot>
  <PhotoFileShareRoot>\\fredrik_asus\videocatalog\photos</PhotoFileShareRoot>
  <ActorPhotoFileLocalRoot>c:\videocatalog\actorphotos\</ActorPhotoFileLocalRoot>
  <ActorPhotoFileShareRoot>\\fredrik_asus\videocatalog\actorphotos</ActorPhotoFileShareRoot>
  <NetworkPort>8754</NetworkPort>
  <AllowRemoteConnections>true</AllowRemoteConnections>
  <ServerUploadFolder>c:\videocatalog\uploads</ServerUploadFolder>
</ServerConfiguration>