VDON Call

VDON Call is a WebRTC/NDI-based remote caller ingest solution for live video productions, based on the two swiss army knifes in their field: the awesome, low-latency, P2P-based video streaming facility VDO.Ninja, and the awesome, ultra-flexible video mixing software OBS Studio. An instance of the landing page, the corresponding URL trampoline, of VDON Call is publically available under the meaningful name URL caller.studio (caller of the studio).

Motivation

The name VDON Call (spoken ViDeO Ninja Call) is derived from vMix Call, the name of the great but weaker product VDON Call is intended to substitute in practice, if necessary. The motivation and primary killer feature of VDON Call, in contrast to the otherwise also awesome vMix Call, is that the callers can ingest their camera and content in parallel and without having to fiddle around with two independent Browser tabs and that this dual-ingest approach is available for 8 callers in total instead of just 4 (because vMix Call supports a maximum of 8 incoming video streams only). The resulting ingested video streams are available as NDI streams on the local network and can be easily consumed (usually on a separate computer) with a video mixing software like vMix or OBS Studio.

Constraints

The underlying VDO.Ninja parameterization is very opinionated and heavily optimized for a total of 16+1 video streams, to allow one to ingest up to 8 callers with their camera and shared content (screen or window) into the video production and to provide a return feed from the video production back to the callers. The callers can hear but not see each other, except the provided return feed is some sort of multiview and shows the callers. This is intentional, as this is not a general video conferencing solution, but a dedicated way to ingest remote video callers of a production. Please keep this in mind.

There are two particular bottlenecks in this approach, which are addressed by the particular VDO.Ninja parameterization:

1. CALLER INGEST FEEDS:

   The caller receiver-side is an OBS Studio, acting as a WebRTC-to-NDI gateway and has to run on a medium gaming PC (Intel Core i7 11700KF, NVIDIA RTX 3070Ti, 32GB RAM) with just about a maximum of about 80% overall resource utilization. As a result, the callers camera use 720p24 and a bitrate of 2000 kbps with H.264 and the callers content use 1080p6 and a bitrate of 3000 kbps with H.264.

2. PRODUCTION RETURN FEED:

   The production sender-side, providing the return feed of the video production back to the caller, should require no more than half of the uplink (50 mbps) and downlink (100 mbps) of the usual Internet connections available in Germany. As a result, the return feed uses 1080p29.97 and a bitrate of 3000 kbps with VP8.

Installation

The solution consists of the following parts which have to be installed:

1. URL TRAMPOLINE:

   This is just the files index.html, vdon.css and vdon-call-logo.svg, which act as a URL trampoline, a tiny webpage which generates the underlying, heavily parameterized VDO.Ninja URLs. Install it as index.html to an arbitrary Internet service. The page without any attached parameters is usually only used by the video production crew to generate URLs. The page with attached parameters is used by all VDON Call components (see below) to expand the short URLs to the underlying VDO.Ninja URLs. For convenience reasons, an instance of this URL trampoline is publically available under the meaningful URL caller.studio.

2. WEBRTC-TO-NDI GATEWAY:

   This is an OBS Studio instance, driven by the files obs-studio/basic.ini, obs-studio.json, obs-studio.lua and obs-studio.html. These are the OBS Studio profile, scene collection, configuration update script and optional entry page, which establish OBS Studio as a WebRTC-to-NDI gateway. For this to work, you have to locally install the following prerequisites:

   • NDI Tools
   • OBS Studio
   • OBS NDI

   Then use Profile → Import for importing the obs-studio/ directory. Then use Scene Collection → Import for importing the obs-studio.json file. Then, use Tools → Scripts → + for importing the obs-studio.lua file.

   Once this is done, use Tools → Scripts → obs-studio.lua to update all browser sources in OBS Studio with your URL trampoline information. The "Base URL" is either https://caller.studio/, file://<path-to-obs-studio.html>/#https://caller.studio/ (if you want to use the optional entry page -- usually just for debugging purposes), or your own URL where you deployed the URL trampoline (see above).

3. VIDEO PRODUCTION:

   This is the actual video production which requires a video mixing software like vMix or OBS Studio to consume and mix the NDI streams of the callers and produce a "program" (aka output) which can act as the return feed video stream to the callers. This really has to run on a computer which is separate (but in the same LAN) to the WebRTC-to-NDI gateway.

Components

The solution consists of the following logical parts under run-time:

• URL Trampoline

  This is the URL trampoline which is deployed to a publically reachable URL like https://caller.studio, and which can be used by the production crew to easily generate the required "short" URLs based on unique organisation name and a room name (unique per organisation). Its trampoline functionality on access of these "short" URLs expand to the underlying, technical, heavily-parameterized VDO.Ninja URLs.

  

• Caller (Sender)

  This is the caller sender-side, i.e., Google Chrome opened a URL like https://caller.studio/#/example/example/C1. The result the VDO.Ninja user interface, configured for a caller participating in a room where the user can only see its own camera and shared content preview and the return feed of the production (here a multiview from vMix). You can also attach the following optional parameters to the URL (concatenate multiple ones with "&"):

  • vdon=<key>:<val>,... (apply arbitrary VDO.Ninja options)
  • version=<id> (force use of a particular VDO.Ninja version: v20, v21, v22, dev, alpha, beta, or even fully-qualified URL to own VDO.Ninja instance)
  • quality=high (use 30/30 FPS and 3/3 Mbps instead of 24/6 FPS and 2/2 Mbps for camera/screen)
  • codec=<id> (use codec x264, vp9, av1 or webp instead of vp8)
  • audiocodec=<id>(use codec pcm, or g722 instead of opus)
  • cam=<name> (camera device)
  • mic=<name> (microphone device)
  • spk=<name> (speaker device)
  • lang=<cc> (ISO country code for UI language)
  • bg=<type> (background replacement: blur, green or canvas)
  • ducking (use audio ducking/noisegate)
  • audiodelay=<ms> (delay audio)
  • videodelay=<ms> (delay video) An example sender URL would be https://caller.studio/#/example/example/C1?cam=Standard&mic=Standard&lang=de&bg=blur

  

• Caller (Receiver)

  This is the heart of the solution, the caller receiver-side, i.e., OBS Studio with the VDON Call profile and scene collection loaded which accesses URLs like https://caller.studio/#/example/example/C1?stream=camera and https://caller.studio/#/example/example/C1?stream=content. For an overview, a special multiview scene is in preview (and also exposed via NDI for inclusion into the multiview of the production -- see below). You can also attach the following optional parameters to the URL (concatenate multiple ones with "&"):

  • vdon=<key>:<val>,... (apply arbitrary VDO.Ninja options)
  • version=<id> (force use of a particular VDO.Ninja version: v20, v21, v22, dev, alpha, beta, or even fully-qualified URL to own VDO.Ninja instance)
  • quality=high (use 30/30 FPS and 3/3 Mbps instead of 24/6 FPS and 2/2 Mbps for camera/screen)
  • codec=<id> (use codec x264, vp9, av1 or webp instead of vp8)

  

• Production (Sender)

  This is the video production. It uses a video mixing software like vMix (shown here) or OBS Studio (as a decent alternative). This production receives all callers (and the multiview) from OBS Studio via NDI and mixes them into a resulting program video stream. For convenience reasons, beside this program video stream, this vMix instance also exports a multiview video stream which shows the caller ingest multiview, selected callers, the program video stream and the clock. Either the program or (as shown shere) the multiview video stream is then send back to the callers via a companion Google Chrome instance accessing the URL https://caller.studio/#/example/example/A1 (standard) or https://caller.studio/#/example/example/A2?mode=meshcast (using the VDO.Ninja meshcast feature). Similar to the mode=meshcast, you can attach the following optional parameters to the URL (concatenate multiple ones with "&"):

  • vdon=<key>:<val>,... (apply arbitrary VDO.Ninja options)
  • version=<id> (force use of a particular VDO.Ninja version: v20, v21, v22, dev, alpha, beta, or even fully-qualified URL to own VDO.Ninja instance)
  • quality=high (use 30/30 FPS and 3/3 Mbps instead of 24/6 FPS and 2/2 Mbps for camera/screen)
  • codec=<id> (use codec x264, vp9, av1 or webp instead of vp8)
  • cam=<name> (camera device)
  • mic=<name> (microphone device)
  • spk=<name> (speaker device) An example URL would be https://caller.studio/#/example/example/A1?cam=vMix Video External 2&mic=Line 2 (Virtual Audio Cable)

  

• Director (Control)

  This is the optional VDO.Ninja director view, i.e., Google Chrome accessing the URL https://caller.studio/#/example/example/D1 where the production crew can send messages to the callers, adjust the parameters of the callers, etc. The director view intentionally has all incoming video/audio feeds disabled and also does not send an own outgoing video/audio feed to not cause any additional network traffic on the callers. You can also attach the following optional parameters to the URL (concatenate multiple ones with "&"):

  • vdon=<key>:<val>,... (apply arbitrary VDO.Ninja options)
  • version=<id> (force use of a particular VDO.Ninja version: v20, v21, v22, dev, alpha, beta, or even fully-qualified URL to own VDO.Ninja instance)

  

Design Cruxes

The design of VDON Call is based on the following cruxes:

• OPTIMIZED VIDEO ENCODINGS: The Caller Ingest Feeds use the video codec H.264, reduced resolutions (720p and 1080p), and reduced framerates (24 and 6) to allow the WebRTC-to-NDI gateway to cope with the decoding of 16 video streams and the re-encoding of 16+1 video streams. In contrast to this, the Production Return Feed uses the video codec VP8 as it provides better overall quality (H.264 on smaller sizes regularly degrades the video stream to be greyed or washed out).

• MINIMUM TOTAL VIDEO STREAMS: The Director mode in VDO.Ninja is forced to not use any video or audio at all, to reduce the minimum number of total video streams in the room and especially reduce the load on the callers as good as possible.

• REDUCED USER INTERFAC CLUTTER: The Caller mode in VDO.Ninja is forced to be as simple as possible from a user interface perspective (many features are intentionally disabled) to not confuse the callers as good as possible.

• SIMPLE AND INTUITIVE URLS: The Caller URLs use the meaningful "caller.studio" domain and are reduced to the absolute minimum of the required information. All technical aspects of the underlying VDO.Ninja facility are hidden and are provided on-the-fly by expanding the URLs internally.

• NAMESPACE CONFLICT AVOIDANCE: The VDO.Ninja room identifiers and stream identifiers are based on the hashing of an organization and room name. This way VDON Call does not "trash" the namespace of VDO.Ninja too much and is less likely in conflict with regular VDO.Ninja uses.

• USER INTERFACE STYLING: The VDO.Ninja user interface, especially the entry page for selecting devices and the chat module, in dark mode and by default look rather unpolished. For an improved user acceptance, the appearance of the user interface of VDO.Ninja is improved with a VDON Call specific style.

Simplified Usage

For just ingesting 1-4 remote guests with VDON Call instead of vMix Call into vMix, you can alternatively directly run the receiver side inside a Web Browser type input of vMix (for vMix 25 it is an ancient Chromium V86, but for vMix 26 it is at least V103).

For ingesting a guest with 720p24 (camera) and 1080p6 (content) use:

• Guest URL: https://caller.studio/#/example/example/C1
• vMix Input 1 URL: https://caller.studio/#/example/example/C1?stream=camera
• vMix Input 2 URL: https://caller.studio/#/example/example/C1?stream=content

For ingesting a guest with 1080p30 (camera) and 1080p30 (content) use:

• Guest URL: https://caller.studio/#/example/example/C1?quality=high
• vMix Input 1 URL: https://caller.studio/#/example/example/C1?quality=high&stream=camera
• vMix Input 2 URL: https://caller.studio/#/example/example/C1?quality=high&stream=content

See Also

This solution was announced 2022/03 in the following communities:

• Reddit: VDO.Ninja
• Reddit: OBS Studio
• Reddit: vMix
• vMix: vMix Call
• Twitter: Engelschall

Copyright & License

Copyright © 2022-2023 Dr. Ralf S. Engelschall
Licensed under MIT