1. What is SRT?
SRT stands for Secure Reliable Transport. The protocol is built for live video over networks where delay, jitter or packet loss can happen. Instead of just accepting lost packets, SRT can request retransmissions within the latency window you set.
| Feature | In short |
|---|---|
| Name | Secure Reliable Transport |
| Purpose | Live video over networks with delay, jitter or packet loss |
| Lost packets | can be requested again within the chosen latency window |
| Network paths | one path; for several paths there is SRTLA |
| Core features | encryption and packet retransmission, as described in the SRT project documentation |
This page deliberately covers transport over a single network path: from the encoder to a reachable SRT receiver and on to OBS. That is not the same as bonding. If you want your stream to keep running over a second connection when one line drops, read the SRTLA & bonding guide next.
Technical background: the official SRT project documentation describes encryption and packet retransmission as core features. Whether a lost packet still arrives in time depends on the network and on the latency window you set.
2. Understanding RTT and latency in SRT
RTT stands for round trip time. It is the time a data packet needs to travel from the sender to the receiver and for a reply to come back. It is usually given in milliseconds. So RTT is a there-and-back measurement and not automatically the same as the delay you see in the live stream.
| RTT | What it means | Practical consequence |
|---|---|---|
| low and stable | Replies come back quickly and evenly | a smaller latency window may be possible |
| high but stable | long distance or a slow network path | plan for more buffer |
| fluctuating heavily | jitter, cell changes or an overloaded path | add headroom and investigate the connection |
| high together with packet loss | critical transmission path | lower the bitrate or use a different network path |
A single high RTT reading does not automatically mean a dropout. It gets critical when RTT, jitter and packet loss keep climbing and the SRT latency window no longer leaves enough time for retransmissions.
No fixed ideal number: Route, mobile network, relay location and SRT latency all affect how you read it. What matters is whether RTT stays stable over a longer period and the stream arrives without growing packet loss.
SRT latency is a buffer for network transport, not the total delay from camera to viewer. A window that is too tight leaves lost packets little time to catch up; a larger window adds transport delay. The SRT documentation on latency explains this difference in detail.
3. When SRT is enough and when you need SRTLA
SRT is the right starting point when a single network path has enough upload headroom and you need stable transport to OBS or a relay. SRTLA extends that SRT-based path with several connections used at the same time and a compatible counterpart. It is the next step when your stream should not hang on a single line.
Here we stick to setting up and diagnosing SRT over one path. Bonding has its own page, so each guide keeps a clear job.
4. SRT in IRL streaming
In mobile streaming, the encoder can send the stream over SRT to a relay server. From there the signal goes on to OBS, a decoder or other streaming infrastructure, for example.
| Area | In practice |
|---|---|
| Latency | A larger buffer gives retransmissions more time but adds to the end-to-end delay. |
| Packet loss | SRT can have lost packets retransmitted within the available time window. |
| Network | Especially interesting for mobile data, Wi-Fi, internet uplinks and remote production. |
For a first test, a compatible sender and a reachable SRT counterpart are enough. You don't need a second modem. What matters most is a stable upload, the right bitrate, and matching publish and play IDs with the same personal key.
5. Testing an SRT connection in practice
A green connection status alone doesn't prove your picture will stay stable out there. Check the whole path with exactly the network, bitrate and receiver you plan to use later:
- Check the network path: Look at upload speed and fluctuations at the planned location; there has to be headroom above your chosen video bitrate.
- Connect sender and receiver: Enter the SRT target, UDP port and stream ID to match the server; the example addresses are further down.
- Check the picture in OBS: Don't just look for "connected." Watch for smooth video and clean audio over several minutes.
- Watch the numbers: Judge RTT, packet loss and retransmissions together. A single short spike says less than a bad trend that lasts.
- Retest on the move: If the network fluctuates, check upload bitrate and latency window first. If your only path drops out completely, SRT can't replace it with a second one.
6. Setting up your own SRT server automatically
The glowf1sh SRTLA receiver isn't only for bonded SRTLA connections. The Docker container also includes a regular SRT live server. That lets you send a stream over a single internet connection to the server via SRT and pull it back into OBS.
One installer for your SRT server: The tested automatic installer sets up the receiver on a suitable Debian or Ubuntu server. Although the container also handles SRTLA, you can use it here purely as an SRT server for a single network path. After installation, you run the status check and connect the sender and OBS as shown below.
Cost and license: According to the vendor information on the CC-BY-NC image, there is no license fee for the container for private, non-commercial personal use. That doesn't make your own SRT server free overall: a virtual server or hardware, power, internet and maintenance still cost money. You may not offer a paid hosting or rental service with this image without additional permission.
| Use | Address and stream ID |
|---|---|
| Send SRT to the server | srt://YOUR-SERVER-IP:8282publish/live/YOUR_LONG_KEY |
| Receive SRT in OBS | srt://YOUR-SERVER-IP:8282?streamid=play/live/YOUR_LONG_KEY |
| Check status | http://YOUR-SERVER-IP:8181/stats |
One container, both paths: For regular SRT you use UDP port 8282. If you later want to bond several LTE, 5G or Wi-Fi connections, the SRTLA input is available as well.
The installer handles the server setup; you enter your personal stream details in the sender and in OBS afterwards. The guide also covers port forwarding and the check after a restart.
7. Narrowing down common SRT problems first
| Symptom | Check first |
|---|---|
| No picture in OBS | Is the server reachable? Are the UDP port, SRT address and play/live/... stream ID correct? |
| Picture stutters or shows errors | Are packet loss or RTT rising? Does your upload leave headroom for the set bitrate? |
| Picture arrives, but clearly too late | Check SRT latency, extra buffers and the rest of the processing chain separately. |
| Stream drops when the network fails | A single SRT path has no backup path; look at SRTLA for several paths. |
When testing, change only one setting at a time and watch the trend afterward. Otherwise it's hard to tell whether bitrate, buffer or network path was the cause.
8. SRT or RTMP?
SRT is especially interesting for transport over problematic networks. RTMP/RTMPS, on the other hand, is still often the last leg from OBS or an encoder to the streaming platform. Both protocols can therefore do different jobs in the same workflow.
9. SRT FAQ
What does SRT mean in streaming?
SRT stands for Secure Reliable Transport. It is built for live video over networks where delay, jitter or packet loss can happen. Within the latency window, SRT can ask for lost packets again.
Does SRT replace bonding with multiple connections?
No. SRT runs over a single network path. If that path drops out completely, no second path takes over. To use several paths at the same time, you need SRTLA.
Is the SRT latency my stream's total delay?
No. SRT latency is a buffer for transport across the network, not the whole delay from camera to viewer. A larger window gives retransmissions more time but adds transport delay.
What is a good RTT for SRT?
There is no fixed ideal number. Check whether RTT stays stable over a longer period and whether packet loss stays quiet while it does. A single spike tells you little, a bad trend that lasts tells you a lot. You can run numbers with the latency calculator.
Which port does the SRT server from the guide use?
In the setup described here, using the glowf1sh receiver, regular SRT uses UDP port 8282. The example addresses are in the section on your own SRT server. With any other server, follow what the operator specifies.