Integrations

Connecting real-world encoders, decoders, and software to Punch.

Overview

Punch generates ready-to-use connection strings for every major SRT implementation. Once a session is created and both peers have registered, Punch provides the exact command or configuration needed — no manual parameter assembly.

This guide covers manual configuration for each platform. In practice, the Punch web UI generates these automatically.

Software

FFmpeg

FFmpeg has native SRT support via libsrt. It is the most flexible tool for SRT workflows.

Send (encoder side):

ffmpeg -re -i input.ts -c copy -f mpegts \
  "srt://PEER_IP:9000?mode=rendezvous&latency=200000&passphrase=PUNCH_PASSPHRASE&pkt_size=1316"

Receive (decoder side):

ffmpeg -i "srt://PEER_IP:9000?mode=rendezvous&latency=200000&passphrase=PUNCH_PASSPHRASE" \
  -c copy output.ts

Critical notes:

Parameter	Value	Common mistake
`latency`	In microseconds	Using milliseconds (200 instead of 200000)
`mode`	`rendezvous`	Omitting mode (defaults to caller)
`pkt_size`	`1316`	Using other values corrupts MPEG-TS
`passphrase`	10-79 chars	Shell quoting issues with special characters

Note on pkt_size versus payloadsize: libsrt accepts both names as aliases for the same parameter. Punch’s auto-generated strings use pkt_size because it is the canonical name in current libsrt and is accepted by every tool tested (FFmpeg, GStreamer, srt-live-transmit). Older guides that show payloadsize are not wrong — but pkt_size is what Punch emits and what its test suite asserts on.

Encoding pipeline example (1080p50 H.264):

ffmpeg -f v4l2 -framerate 50 -video_size 1920x1080 -i /dev/video0 \
  -f alsa -i hw:0 \
  -c:v libx264 -preset ultrafast -tune zerolatency -b:v 8M \
  -c:a aac -b:a 128k \
  -f mpegts \
  "srt://PEER_IP:9000?mode=rendezvous&latency=200000&passphrase=PUNCH_PASSPHRASE"

Low-latency tips:

-preset ultrafast -tune zerolatency — minimises encoding latency
-g 50 — keyframe interval matching framerate (1 second)
-flags +cgop — closed GOP for better error resilience
Consider -x264-params "sliced-threads=1" to reduce per-frame latency

OBS Studio

OBS has native SRT support since version 27.

Settings → Stream:

Field	Value
Service	Custom…
Server	`srt://PEER_IP:9000?mode=rendezvous&latency=200000&passphrase=PUNCH_PASSPHRASE`
Stream Key	(leave empty)

Settings → Output → Streaming:

Field	Recommended
Encoder	x264 (or hardware: NVENC, QSV, AMF)
Rate Control	CBR
Bitrate	6000-10000 kbps
Keyframe Interval	1s or 2s
Tune	zerolatency

OBS SRT output limitations:

OBS only supports SRT as output (streaming), not as input (source)
For SRT input in OBS, use the Media Source with an srt:// URL
Custom OBS builds (obs-srt) may offer additional SRT features

vMix

vMix has comprehensive SRT support with multi-audio capabilities.

Add SRT output:

Settings → Outputs → Gear icon
Select “SRT” as output type
Configure:

Field	Value
Hostname	`PEER_IP`
Port	`9000`
Type	Rendezvous
Latency	`200` (vMix uses milliseconds)
Passphrase	`PUNCH_PASSPHRASE`

Add SRT input:

Add Input → Stream/SRT
Enter the srt:// URL

vMix MABC (Multi Audio Bitrate Channel):

vMix can output 4 stereo audio pairs on separate MPEG-TS PIDs:

PID	Content
Audio 1	Programme mix
Audio 2	Talent microphones
Audio 3	Ambient/effects
Audio 4	IFB/intercom

Note: Not all decoders read multiple audio PIDs. Test your receive chain.

VLC

VLC can play SRT streams but has limitations.

Play an SRT stream:

Media → Open Network Stream → srt://PEER_IP:9000?passphrase=PUNCH_PASSPHRASE

Limitations:

VLC is always SRT Caller via the GUI — it cannot act as a Listener

For rendezvous mode, use VLC’s command line:

vlc "srt://PEER_IP:9000?mode=rendezvous&passphrase=PUNCH_PASSPHRASE"

VLC adds significant decode buffering — not suitable for ultra-low-latency monitoring

GStreamer

GStreamer supports SRT via the srtsrc and srtsink elements.

Receive:

gst-launch-1.0 \
  srtsrc uri="srt://PEER_IP:9000" mode=rendezvous latency=200 passphrase=PUNCH_PASSPHRASE \
  ! tsdemux \
  ! decodebin \
  ! autovideosink

Send:

gst-launch-1.0 \
  v4l2src ! videoconvert ! x264enc tune=zerolatency bitrate=8000 \
  ! mpegtsmux \
  ! srtsink uri="srt://PEER_IP:9000" mode=rendezvous latency=200 passphrase=PUNCH_PASSPHRASE

Critical gotcha: GStreamer silently ignores passphrases shorter than 10 characters. The connection appears to succeed but is unencrypted. Always use passphrases of 10+ characters. Punch generates 32-character passphrases by default, avoiding this entirely.

CasparCG

CasparCG has no native SRT support. There are two workarounds:

Option 1: srt-live-transmit relay (recommended)

Run a local relay that converts between SRT and UDP:

# Receive SRT and output to local UDP (for CasparCG input)
srt-live-transmit \
  "srt://PEER_IP:9000?mode=rendezvous&passphrase=PUNCH_PASSPHRASE&latency=200000" \
  "udp://127.0.0.1:5000" -v

Then configure CasparCG to use udp://127.0.0.1:5000 as input.

Option 2: Custom FFmpeg build

Build FFmpeg with --enable-libsrt and use it as CasparCG’s FFmpeg backend. This requires modifying CasparCG’s build configuration.

Note: CasparCG GitHub issues #1251 and #1535 requesting native SRT support remain open.

srt-live-transmit

The reference SRT relay tool from Haivision. Useful for protocol bridging and relay setups.

# SRT rendezvous → SRT listener (relay to studio)
srt-live-transmit \
  "srt://PEER_IP:9000?mode=rendezvous&passphrase=PUNCH_PASSPHRASE&latency=200000" \
  "srt://:10000?mode=listener&passphrase=STUDIO_KEY" -v

# SRT rendezvous → UDP multicast (local distribution)
srt-live-transmit \
  "srt://PEER_IP:9000?mode=rendezvous&passphrase=PUNCH_PASSPHRASE&latency=200000" \
  "udp://239.1.1.1:5000?adapter=eth0" -v

Hardware encoders

Magewell Ultra Encode

Feature	Support
SRT Caller	Yes
SRT Listener	Yes
SRT Rendezvous	Yes (firmware 2.x+)
Bonding	No (libsrt bonding)
Multi-audio	Stereo + embedded HDMI audio

Configuration: Web UI → Encoding → Protocol: SRT → Mode: Rendezvous → enter peer IP, port, passphrase, latency.

Teradek (Vidiu, Cube, Prism)

Feature	Support
SRT Caller	Yes
SRT Listener	Yes (Cube, Prism)
SRT Rendezvous	No
Bonding	Via Sharelink (proprietary cloud)

Limitation: Teradek does not support SRT rendezvous mode. To use with Punch, the Teradek must connect as SRT Caller to an srt-live-transmit relay that converts to rendezvous.

Kiloview (N-series)

Feature	Support
SRT Caller	Yes
SRT Listener	Yes
SRT Rendezvous	Model-dependent (check firmware)
NDI bridge	Yes (N-series)

Haivision (KB, Makito)

Haivision created SRT — their encoders have the most complete implementation.

Feature	Support
SRT Caller	Yes
SRT Listener	Yes
SRT Rendezvous	Yes
Bonding	Via SRT Gateway (proprietary)
Multi-audio	Up to 8 stereo pairs

Blackmagic (ATEM Mini Pro, Web Presenter)

Feature	Support
SRT Caller	Yes
SRT Listener	No
SRT Rendezvous	No

Critical limitation: Blackmagic devices are SRT Caller only. They cannot be a Listener or use Rendezvous mode. To use with Punch, the receive side must run an srt-live-transmit relay in Listener mode.

AJA HELO

Feature	Support
SRT Caller	Yes
SRT Listener	Yes
SRT Rendezvous	Firmware-dependent

Multi-camera ISO setup

For a multi-camera ISO production with Punch:

Session: "show-2026-03-06"
Streams: cam-wide, cam-close, cam-roving, pgm-return

Camera 1 (wide)    → punch.thåst.se/s/show-2026-03-06 → scan QR → claim "cam-wide"
Camera 2 (close)   → punch.thåst.se/s/show-2026-03-06 → scan QR → claim "cam-close"
Camera 3 (roving)  → punch.thåst.se/s/show-2026-03-06 → scan QR → claim "cam-roving"
MCR return feed    → punch.thåst.se/s/show-2026-03-06 → scan QR → claim "pgm-return"

Each camera gets:

Its own SRT rendezvous connection
A unique stream slot in the session
Individual health monitoring on the dashboard
Tally state (program/preview/off)

All coordinated through one session URL.

Audio channel mapping

Scenario	Approach	MPEG-TS PIDs
Simple stereo	Single audio PID	1
Programme + IFB	Two audio PIDs	2
Full ISO (per-camera mix)	One audio PID per camera stream	1 per stream
vMix MABC	Four stereo PIDs in one stream	4
Discrete channels	AAC 5.1/7.1 in single PID	1

For discrete multi-channel audio, AAC is the most compatible codec. Opus offers lower latency but has less universal decoder support in broadcast hardware.