VRChat YouTube Search API

An API server for building YouTube search UIs in VRChat, with pagination, thumbnails and channel icons.

Test in browser: https://api.u2b.cx/test.html

Endpoints

GET /search

Get YouTube videos for a search query.

Required query parameters

• pool: id of the VRCUrl pool, only letters numbers hyphens or underscores, optionally followed by an integer for pool size.
• input: youtube search query. All chars up to and including this exact unicode char → are ignored, and then whitespace is trimmed. THIS MUST BE THE LAST QUERY PARAMETER AS ALL CHARS AFTER IT ARE CAPTURED VERBATIM (so you can type & etc without encoding). If this contains a url with a playlist ID, playlist results will be loaded instead (which is a bit different, up to 100 results and no descriptions).

Optional query parameters

• thumbnails: set to 1, true, yes, on or whatever to load thumbnails
• icons: set to 1, true, yes, on or whatever to load channel icons
• mode: If set to latestontop will search without filter to include the "Latest from ..." section (if it exists). Otherwise it uses search filter for Videos only sorted by Relevance.
• bp: Custom YouTube search filter. Go to youtube search site and set some filters and you will see the corresponding bp value in the URL. Overrides mode option. Use at your own risk (experimental).

Example URL

https://api.u2b.cx/search?pool=example10000&input=   Type YouTube search query here →                       penile apparatus

Response format

JSON object:

• results: Array of Object
  • vrcurl: (integer) index of VRCUrl that will redirect to the youtube url, or serve JSON with captions if used with string loader.
  • live: (boolean) whether it's a live stream
  • title: (string) i.e. "Nyan Cat! [Official]"
  • id: (string) YouTube video id i.e. "2yJgwwDcgV8"
  • description?: (string) short truncated description snippet i.e. "http://nyan.cat/ Original song : http://momolabo.lolipop.jp/nyancatsong/Nyan/" (playlist results don't have this)
  • lengthText?: (string) i.e. "3:37"
  • longLengthText?: (string) i.e. "3 minutes, 37 seconds"
  • viewCountText?: (string) i.e. "2,552,243 views" or "575 watching" for live streams (playlist results don't have this)
  • shortViewCountText?: (string) i.e. "2.5M views" (streams don't have this)
  • uploaded: (string) i.e. "12 years ago"
  • channel: (object)
    • name: (string) i.e. "NyanCat"
    • id: (string) i.e. "UCsW85RAS2_Twg_lEPyv7G8A"
    • icon?: (object) (playlist results don't have this)
      • x: (integer) px from left
      • y: (integer) px from top
      • width: (integer)
      • height: (integer)
  • thumbnail?: (object)
    • x: (integer) px from left
    • y: (integer) px from top
    • width: (integer)
    • height: (integer)
• imagesheet_vrcurl?: (integer) index of the vrcurl for the collage of thumbnails and/or icons
• nextpage_vrcurl: (integer) index of the vrcurl that will serve the JSON for the next page of results

GET /trending

Gets Trending YouTube videos. Identical to /search but without input parameter, and response includes additional field:

• tabs: Array of Object
  • name: (string) Tab title ("Now", "Music", "Gaming", "Movies")
  • vrcurl: (integer) index of vrcurl to load that tab (same response format)

GET /vrcurl/{pool}/{index}

• {pool}: must be same as pool param in search endpoint.
• {index}: vrcurl index number

Response may be 302 redirect to youtube url, image/png for imagesheet, application/json for next page (see response format above) or trending tab or video json data (see below).

Video metadata JSON format

• captions: Array of Object
  • name: (string) caption track name like "English" or "English (auto-generated)"
  • id: (string) id like .en or a.en
  • lines: Array of Object
    • start: (float) video seconds when the caption is displayed
    • dur: (float) seconds to display the caption
    • text: (string) caption text

VRCUrls

Since VRCUrls are immutable you must create a pool of them which the server will correspond with to receive user selections. Create an array of 10,000 VRCUrls like so:

VRCUrl[] vrcurl_pool = [
	new VRCUrl("https://api.u2b.cx/vrcurl/{pool}/0"),
	new VRCUrl("https://api.u2b.cx/vrcurl/{pool}/1"),
	new VRCUrl("https://api.u2b.cx/vrcurl/{pool}/2"),
	// etc...
]
//todo: provide tool to auto generate

{pool} must be a unique string in the format ^[a-z-_]+\d*$. You can specify the pool size by suffixing with an integer, or else the default is 10,000.

All resources (youtube urls etc) referenced in the search results will be substituted by an integer that is the index of the VRCUrl in this array that will serve the resource.

Imagesheet

Video thumbnails and channel icons are collated together into one image and served at a VRCUrl to be loaded by VRCImageDownloader.

Use the x, y, width and height values from the json to crop the image from the sheet. Do not make any assumptions about these values as the server could arrange the images wherever it wants.