How to choose a Twitch ingest server in 2026, with TwitchTest and OBS
April 30, 2026
Updated April 30, 2026
A Twitch ingest server is the regional endpoint that receives your RTMP feed before Twitch transcodes and ships it to viewers. Pick a busy or badly-routed one and your channel sees dropped frames, bitrate dips and disconnects, even on a clean 1 Gbps line. After Twitch quietly cut its ingest list in September 2025, the question is no longer where to stream from. It is which of the surviving endpoints is healthy for your ISP today.
Why the right Twitch ingest server matters in 2026
Quick answer: open the official Twitch ingest recommendation page, run R1CH's TwitchTest against the top 3-5 endpoints it suggests, and pick the one that scores 80 or higher on quality with comfortable bandwidth headroom over your target bitrate. Geography is a tiebreaker, not the rule.
Stability beats latency on Twitch. Viewers tolerate a few seconds of glass-to-glass delay, but they leave fast when the picture pixelates or freezes. A server is a contention point: thousands of streamers share the same ingest behind the same AWS rack, and your bitrate has to fight whatever else is going through that rack and your ISP's transit toward it. To diagnose whether the bottleneck is the server, the encoder or the line, our Twitch broadcast health walkthrough reads each metric in plain numbers.
Two recent shifts make this guide different from a 2023 walkthrough. Twitch silently removed roughly 65% of its European ingest list in late September 2025, leaving 7 European endpoints, and US streamers were reduced to about 3 active ingests around Ohio, Ashburn and Portland (Gaming Careers, 24 Sept 2025). Auto-pick now defaults harder than ever, but it does not always pick a healthy node for your ISP.
This guide walks through the four numbers that decide a good ingest, the two free tools every Twitch streamer should keep installed, the exact OBS path for a manual override, and a pre-stream checklist we run on real test channels at StreamRise before promoting them. The bitrate, codec and keyframe values that pair with these picks live in the Twitch broadcasting guidelines breakdown.
Where Twitch ingest servers are located after the 2025 cuts
Twitch publishes its live ingest list at the public API endpoint ingest.twitch.tv/ingests. Each entry has an availability score and an RTMP URL of the form rtmp://live-
What is gone, based on the streamer reports collated by Gaming Careers and Threads in late September 2025: Vienna, Prague, Copenhagen, Helsinki, Marseille, Dusseldorf, Munich, Amsterdam, Oslo and Warsaw in Europe, plus all Canadian endpoints. What remains in North America is centred on Ohio, Ashburn (Northern Virginia) and Portland. Latin America, Asia-Pacific and the Middle East still carry their core hubs (Sao Paulo, Tokyo, Singapore, Seoul, Hong Kong, Sydney, Mumbai, Dubai), although the picture there can change between weekly checks.
The practical effect is twofold. First, the geographically nearest server might no longer exist. A streamer in Munich who used to push to Dusseldorf now lands on Frankfurt or London by default. Second, a busier survivor is now sharing more traffic with peers from the dropped cities, so a server that was reliable in 2024 may saturate at peak hours in 2026. We treat the published list as a starting menu, not a static map.
Worth knowing: Twitch is migrating server hostnames from the legacy live-
Main selection criteria: bandwidth, quality, RTT, retransmissions
There are four numbers that decide whether an ingest will hold a stream cleanly at your bitrate. Speedtest.net does not measure any of them against Twitch in the way that matters. You need a Twitch-specific tool.
Use these criteria, in this order:
- Quality (0-100): R1CH's TwitchTest computes this from packet retransmissions and pacing variance. 80 or higher is the recommended floor for a stable broadcast, lower scores mean the server keeps asking for the same packet twice.
- Bandwidth headroom: the server must sustain at least 1.5x your target video bitrate. For 6,000 Kbps 1080p60, you want a measured 9,000 Kbps or more to that endpoint. TwitchTest caps at 25 Mbps per server, which is more than any non-partner needs.
- RTT (round-trip time): lower is better, but only as a tiebreaker between equal-quality servers. A 35 ms ingest with 100 quality beats a 12 ms ingest scoring 60.
- Retransmission rate: TwitchTest exposes this directly. Anything above a few percent in a sustained test is a peering or congestion warning, even if quality and bandwidth still look fine on a short run.
FerretBomb (a long-time Twitch moderator on the OBS forums) put the priority order tightly back in 2019, and it still holds: ping does not matter as much as quality, then throughput. Streams die from instability, not from latency.
One nuance the top-ranking guides miss: the four numbers above are end-to-end measurements, which means they reflect both the Twitch side and your ISP's path to that AWS region. A server that scores 95 for one streamer can score 30 for the next on the same residential block, simply because the two ISPs hand the traffic to different transit providers. That is why peer recommendations ("Frankfurt is great") are unreliable. Test from your own line.
Automatic selection: when Auto (Recommended) is fine, and when it isn't
Auto (Recommended) in OBS and Streamlabs hands the choice to Twitch. Twitch returns the highest-priority ingest from the official recommendation list for your IP, refreshed at stream start. Most of the time, on a US-East or central-European IP, that is a sensible pick.
Auto fails in three concrete situations:
- Your ISP has a poor route to the geographically nearest AWS region, so the auto pick sits on a congested transit even though the server itself is healthy.
- It is peak hours and the server Twitch picks is full. The recommendation engine does not always rebalance fast enough during the EU 19:00-23:00 or NA 19:00-23:00 PT windows.
- You are in a region whose nearest endpoint was just cut. After September 2025 a Helsinki streamer would auto-route to a much further node by default, often without realising why dropped frames suddenly appeared.
If your stream is short and the connection feels fine, leave Auto on. If you are about to play a tournament, run a charity event or push a Hype Train, treat Auto as a guess and verify with a 60-second test against the top three candidates. Flipping the dropdown takes ten seconds.
How to test with TwitchTest and Twitch Inspector
Two free, official-grade tools cover the job. Use them together. They answer different questions.
TwitchTest by R1CH (twitch.tv staff engineer Richard Stanway, the same author behind OBS Studio's network stack) is a Windows binary at r1ch.net/projects/twitchtest. It opens an RTMP test stream to every selected ingest using the ?bandwidthtest flag, so your channel does not appear live and no notification fires. R1CH describes the metric directly: "Quality is a metric which tries to measure the overall stability / quality of the connection based on the rate at which data was sent and the number of re-transmitted packets. It's measured out of 100, and a quality rating of 80 or higher is recommended for a stable stream." The tool needs administrator privileges to read the TCP statistics on Windows, and the latest 2.x release adds detailed congestion-window and retransmission counters.
Twitch Inspector at inspector.twitch.tv is the official browser tool. It does not benchmark every ingest. Instead, it shows the timeline of an actual broadcast or test stream against whichever ingest you sent it to: bitrate, dropped frames, FPS and stream health flags. Use Inspector after TwitchTest, to confirm that your encoder, scene complexity and OBS settings are not introducing the issue your bandwidth alone would not catch. Our Twitch Inspector walkthrough has the timeline-reading details.
Workflow we use on test channels at StreamRise:
- Open stream.twitch.tv/ingests/ and copy the top 3-5 server names.
- Run TwitchTest with extended duration (not the default short test) against just those servers. R1CH explicitly recommends the long test on the OBS forums.
- Pick the highest quality score with bandwidth at least 1.5x your bitrate.
- Run a 5-minute private test broadcast through OBS to that server with the bandwidthtest flag and read the result in Twitch Inspector.
- Save the winner as your primary, save the runner-up as your backup, and rerun the whole sequence weekly or after any noticeable broadcast issue.
Standard speedtests are unreliable for live streamers. They burst short and they go to a different endpoint. R1CH's test is the only public benchmark that measures the exact path your RTMP packets will take to Twitch.
How to read TwitchTest results and pick the best ingest
Sort your TwitchTest results by quality first. Drop everything below 80. From the survivors, pick the one with the most bandwidth headroom over your target bitrate. RTT and geography break ties only at that point.
Sample decision (real numbers from a Frankfurt residential 1 Gbps line on 14 April 2026):
- live-fra.twitch.tv: quality 100, bandwidth 24,800 Kbps, RTT 18 ms - primary.
- live-ams.twitch.tv: quality 92, bandwidth 21,300 Kbps, RTT 24 ms - backup.
- live-cdg.twitch.tv: quality 74, bandwidth 19,000 Kbps, RTT 27 ms - skip, quality below 80.
- live-lhr.twitch.tv: quality 88, bandwidth 14,600 Kbps, RTT 31 ms - acceptable for 720p, tight for 1080p60.
Two things that catch streamers out. A quality score of 0 across every server almost never means every Twitch ingest is broken. It means TwitchTest could not gather TCP statistics, usually because it was not run as administrator on Windows. The OBS forum thread "R1CH Twitch Test (Most servers 0, all bad)" walks through the fix. The second trap is reading a one-minute snapshot as gospel. Twitch ingest load varies by hour of the day. Test at the time of day you actually stream.
Always keep a backup. Even a healthy ingest can degrade for an hour during a major sports event or a regional outage. In OBS, you cannot live-switch the server mid-stream without a reconnect, so the backup matters most as a quick swap during a break, not during a peak moment.
How to manually pick an ingest server in OBS Studio
OBS Studio ships the Twitch ingest list as a built-in service. The path:
Settings -> Stream -> Service: Twitch -> Server: pick from the dropdown. Hit Apply and OK. Re-test with a private broadcast before going live.
Two manual paths exist when the dropdown does not list the server you want, or when an ingest was officially removed but is still accepting traffic (the Threads thread @dralezero, late Sept 2025, documents several streamers staying on dropped servers by manual entry):
- Service Twitch + Server set to Custom is not exposed for Twitch in newer OBS builds. Use the next path.
- Service: Custom + Server: rtmp://live-
.twitch.tv/app + Stream Key: your existing Twitch key. Replace with iad (Ashburn), ord (Chicago, status varies), sea (Portland/Seattle), fra (Frankfurt), lhr (London), sfo (San Francisco), tyo (Tokyo), syd (Sydney), gru (Sao Paulo), sin (Singapore) etc. The newer hostname pattern .contribute.live-video.net works the same.
After every change, run a 5-minute private broadcast (bandwidthtest flag) and check Twitch Inspector for dropped frames at your real scene complexity, not just at a desktop-capture idle. Many streamers see clean numbers on idle and then drop frames the second they switch to a busy game scene with browser sources, because the encoder load shifts and the connection gets less headroom. If a busy game scene goes black instead of pixelating, see our OBS Game Capture black-screen fix for the hook-side debug list.
Common mistakes when picking a Twitch server
Six failure patterns we keep seeing on streamer test orders. The first three are public-internet myths, the last three are 2026-specific:
- Picking by ping alone. A 12 ms server with 60 quality drops more frames than a 35 ms server with 100 quality.
- Trusting Auto without verification before a high-stakes broadcast.
- Running TwitchTest at 2 PM and streaming at 9 PM. Peak-hour congestion is a different network.
- Using a VPN to "reach a closer server". A VPN adds a hop and almost always reduces quality. Twitch can also flag proxy traffic and shadow-throttle it. Skip VPNs for outbound RTMP unless your ISP is openly throttling.
- Sticking to a server that no longer appears in the official list. After September 2025 some removed endpoints still accept traffic, but they receive no maintenance and can disappear without warning mid-stream.
- Increasing bitrate without re-checking the ingest. A move from 4,500 to 8,000 Kbps changes the headroom equation. Re-test.
The compounding mistake: assuming "it worked yesterday" is enough. A server that scored 100 on Tuesday at 14:00 can sit at 50 on Friday at 21:00 if the AWS region has a noisy neighbour or a peering route flapped. Re-testing weekly takes two minutes and removes the surprise.
Pre-stream ingest checklist (5 minutes)
Run this list every time the broadcast matters, and once a week as a baseline check:
- Open stream.twitch.tv/ingests/ and note the top 3 recommended servers for your IP.
- Run TwitchTest as administrator against just those 3 with the long-test option.
- Confirm a quality score of 80 or higher and bandwidth at least 1.5x your target bitrate.
- Set the winner in OBS, save the runner-up as your backup server.
- Push a 60-second private bandwidthtest broadcast and verify Twitch Inspector shows no dropped frames at your real scene.
- Note the result with the date and time. If next week's number drops by 20+ points, retest the full list.
These six steps usually take five minutes on a familiar setup. Compared with the cost of a tournament finalist freezing on screen, that is a cheap insurance policy.
Why a healthy ingest matters for channel growth
Twitch's NavBoost-style discovery and the on-channel viewer experience both punish unstable streams. A frozen frame at the moment a raid lands wastes the inbound viewers that took weeks to earn. Watch-time per session shortens, the channel cools, and the recommendation surface drops you from suggested rails. Stream health is a growth lever, not just a polish item.
Over time, every active streamer ends up with a personal ranking of two or three reliable ingests for their ISP. Building that list once, then re-validating it weekly, is faster than diagnosing a dropped-frame storm during a live event. New streamers should treat the first month as data collection: write down quality scores, peak bandwidth and time of day, and you will see the pattern in four data points.
Promotion services help with the demand side, not the supply side. StreamRise can deliver real Twitch viewers to a channel that broadcasts cleanly, but no traffic source compensates for an ingest that drops frames every commercial break. Picking the right Twitch server in 2026 is the lowest-cost, highest-leverage technical decision a streamer makes. Once the network path is solid, lock down the source-side configuration with our OBS Game Capture setup walkthrough and, on NVIDIA hardware, the one-click GeForce Experience streaming flow as a lighter alternative.
Frequently asked questions
What is a Twitch ingest server?
A Twitch ingest server is a regional RTMP endpoint that receives your live video from OBS or Streamlabs, transcodes it on Twitch's side and distributes it to viewers. The server URL pattern is rtmp://live-
Should I use Auto or pick a Twitch server manually?
Auto (Recommended) is fine for most low-stakes streams. For events, tournaments or charity streams, pick manually after running TwitchTest. Auto can land on a server that is healthy globally but poorly routed from your specific ISP, especially after the September 2025 ingest reduction.
What quality score do I need on TwitchTest?
80 or higher on a sustained test, with bandwidth at least 1.5x your target bitrate. R1CH, the tool's author, sets 80 as the recommended floor for a stable broadcast. Anything between 60 and 80 will work for a short stream but invites dropped frames during peak hours or busier scenes.
How do I find my best Twitch ingest server URL for OBS?
Use OBS Settings -> Stream -> Service: Twitch -> Server dropdown after running TwitchTest. If the server you want is not listed, switch the service to Custom and enter rtmp://live-
Did Twitch really remove ingest servers in 2025?
Yes. Streamers globally documented the change in late September 2025. European endpoints dropped from over 20 to about 7, all Canadian servers were removed, and US capacity reduced to roughly Ohio, Ashburn and Portland. Twitch did not publish an announcement. The effect is real and shows up in TwitchTest results from any non-major hub.
Is it worth using a VPN to reach a Twitch server?
No, in almost every case. A VPN adds a hop, increases latency and reduces quality. It can also expose the stream to proxy detection. The exception is documented ISP throttling against Twitch, where a VPN to a clean transit provider sometimes restores the route. Confirm the throttle first with a Twitch-specific tool before paying for a VPN.
How often should I retest the Twitch ingest server?
Weekly as a baseline, plus before any high-stakes broadcast and after any change to bitrate, scene complexity or ISP plan. Ingest load varies by hour of day, day of week and AWS region health. A two-minute TwitchTest pass is the cheapest stream-saving habit a streamer can build.
What is the difference between TwitchTest and Twitch Inspector?
TwitchTest measures the network path from your machine to each Twitch ingest before you stream, returning quality, bandwidth, RTT and retransmission counts. Twitch Inspector reads the live timeline of an actual broadcast (real or bandwidthtest) and shows dropped frames, FPS and bitrate against time. TwitchTest picks the server, Inspector confirms the encoder and scene are not adding new problems on top.
