Webcasting Crew Positions

Webcasting covers every show that goes out to a stream: conferencing platforms like Zoom and Teams, and production encoders like vMix and OBS that push to YouTube, a CDN or any RTMP or SRT endpoint.

Behind Every Stream

The Stream Doesn't Fail at the Encoder.
It Fails at the Link.

A webcast that buffers is almost never an encoder problem. It is a bandwidth problem that showed up at the encoder. The stream needs a steady share of the venue's upload, and it needs that share to still be there when the client's team starts a Zoom, a laptop syncs its cloud drive, and the hotel's guest network wakes up. Headroom is the ratio between what the link can carry and what the stream asks for. Isolation keeps that headroom from being borrowed. A second path covers the moment the first one disappears. Set up a venue below and see whether the stream holds.

The Venue Link
What the encoder sends. 1080p30 runs 4 to 6 Mbps; 1080p60 runs 8 to 12. Multiple destinations multiply it.
Measured on site, on the actual drop, with a sustained test. The number on the hotel's one-sheet is not this number.
Client laptops, presenter Zooms, registration, guest Wi-Fi, and the one person uploading photos to the cloud during the keynote.
LML Design Choices
What the Viewer Gets
0.0×
Fragile
Headroom: link capacity over stream bitrate
0 Mbps available to the encoder
0 Mbps
Bitrate at VBR peaks
0.0×
Headroom at peaks
Dies
If the primary link drops
3
Single points of failure

Headroom by stream bitrate

Shaded band is the working target, 2× or better. Below 1× the stream cannot be sent at all.

This design Shared link, same venue
Bandwidth headroom versus stream bitrate for this design and a shared venue link
View as table
BitrateThis designShared link
01 / Headroom

Two times the bitrate, sustained

An encoder set to 8 Mbps does not send a flat 8 Mbps. Variable bitrate peaks on motion and scene changes, the protocol retransmits what the network drops, and a second destination doubles everything. The working rule is an upload at least twice the total bitrate, measured as a sustained upload over ten minutes on the actual drop, not a thirty-second speed test from a phone.

Protocol matters too. RTMP rides on TCP and stalls when packets go missing, which is how a brief hiccup becomes a frozen player. SRT and RIST recover lost packets inside a latency budget you choose. LML operators use them where the destination supports it and size the buffer to the link.

02 / Isolation

Nobody borrows the encoder's lane

A venue's upload is shared by everything on its network. The number you measured at 9 a.m. is not the number at 9:05 when doors open. The fix is to stop sharing: the encoder on a hardwired drop, on its own VLAN, with a static address and its traffic prioritized at the switch. Guest Wi-Fi is never an option, and neither is the client's conference-room jack unless the venue IT confirms what else is on it.

This is why the network management seat exists on a webcast crew. Someone owns the switch, the VLANs, the addressing for NDI and Dante, and the conversation with venue IT before load-in, so the stream's lane is reserved before anyone else arrives.

03 / Second path

Plan for the link to disappear

Every single link fails eventually: a venue router reboots, a fiber gets kicked, an ISP has a bad afternoon. A bonded cellular unit, or a second ISP on its own path, carries the same stream in parallel so the loss of the primary is a stutter the viewer may not notice. For shows that cannot go dark, the backup extends to a second encoder and a second stream key.

Then watch the right thing. The encoder's preview tells you what left the building. A phone on cellular, playing the public stream, tells you what the audience sees. The webcast operator watches both, and has the technical-difficulties slate ready before it is needed.

< 1×
Fails

The link cannot carry the stream. Drop the bitrate or find another line.

1 to 1.5×
Fragile

Sends in a quiet room. Buffers the moment anything else moves.

1.5 to 2×
Marginal

Survives steady content. VBR peaks and retransmits eat the rest.

2 to 4×
Solid

The working target. Room for peaks, recovery, and a second destination.

> 4×
Plenty

Enough to raise quality or add outputs. Spend it on redundancy first.

Model notes: bandwidth available to the encoder is the venue upload minus other traffic on the same link, or the full upload when the encoder is isolated on its own VLAN with priority. Headroom is available bandwidth divided by stream bitrate. VBR peaks are taken as 1.5 times the target bitrate. The bonded second path is modeled at 20 Mbps aggregate, a conservative figure for a four-modem unit in a city venue. Single points of failure count the primary link, the venue's ISP, and the encoder, less one for each covered by the second path. Real links vary by hour and by floor; LML measures on site and again at doors.

Need a webcast operator?

Tell us your platform, show date, and what's being streamed. We'll staff it.

Submit a Booking Inquiry