Summary
Through the in-editor MetaHuman Creator plugin, creating a rig or downloading textures for a MetaHuman Character fails with “Server Error” after the HTTP request to Epic’s cloud service times out at a fixed 300 second limit. It reproduces on a fresh project with a default preset on UE 5.6, 5.7 and 5.8. The identical in-editor workflow worked in 5.5.
From my location (Johannesburg, South Africa) on a direct connection, texture download, Joints Only rig and Full Rig all fail this way. Routing through a US-based VPN makes texture download and Joints Only rig succeed, but Full Rig still times out at 300s. This points to two separate problems: an unstable long-haul route from my region to Epic’s us-east-1 services, and a fixed 300s timeout that a legitimately running job cannot fit inside from a higher-latency region. The 300s limit is not client-configurable (see Additional Notes).
Endpoints involved (both us-east-1): mh-uemhc-autorig-service.eeeb.live.use1a.on.epicgames.com and mh-texture-synthesis-service.eeeb.live.use1a.on.epicgames.com
What Type of Bug are you experiencing?
Animation
Steps to Reproduce
On Windows, install the MetaHuman Creator plugin from the Epic Games Launcher (Assets, then Install to a supported engine version) and enable it, along with MetaHuman Live Link.
Create a new project on 5.6, 5.7 or 5.8 and sign in to your Epic account.
Create a MetaHuman Character from a default preset, with no custom mesh.
Attempt any of: Download Textures, create Joints Only rig, create Full Rig.
Observe the request reach 75% / RUNNING, then fail after 300 seconds with “Server Error”. (From a US network route, Download Textures and Joints Only succeed and only Full Rig fails. From a non-US direct connection, all three fail.)
Expected Result
The AutoRig / texture-synthesis service returns the completed data, the MetaHuman Character binds successfully, and the editor continues without an HTTP timeout or Server Error, regardless of the user’s region.
Observed Result
The request is accepted and reaches 75% / RUNNING, then status polling exceeds the fixed 300 second HTTP limit. The editor logs “HTTP request timed out after 300.00 seconds” and “Auto-Rigging of Face failed with code ‘Server Error’” (or “Download of source face/body textures failed with code ‘Server Error’”), then returns the character to the unbound state. When Joints Only succeeds over a US route it completes in 183.971 seconds, under the limit but with little headroom. Screenshots attached.
Affects Versions
5.7
Platform(s)
Windows
Upload an image
Additional Notes
Regression: the same in-editor plugin workflow worked in 5.5. 5.6 introduced the reworked MetaHuman architecture, which is when this started for me. This is not a web-versus-editor difference; I only use the in-editor plugin and have never used the browser-based MetaHuman Creator.
What I tried:
Direct connection (South Africa): all three operations fail at the 300s timeout. Logs show connection instability to Epic’s us-east-1 hosts on this route: “Connection died, retrying a fresh connect”, “client read function EOF fail, only 0/442 of needed bytes read”, “necessary data rewind was not possible”, libcurl error 26. That is a connection dropping mid-transfer, not a clean server rejection.
US-based VPN (US server): texture download and Joints Only rig succeed. Full Rig still times out at 300s. Turning the VPN off brings the failures straight back.
Raising the timeout via a [HTTP] section in the project DefaultEngine.ini (HttpActivityTimeout=1200, HttpReceiveTimeout=1200, HttpConnectionTimeout=1200, HttpTotalTimeout=0) had no effect. With those values set the request still timed out at exactly 300.00 seconds, not 1200. This indicates the 300s limit is enforced on the request inside the MetaHuman plugin and is not read from config, so there is no client-side way to extend it.
Fresh projects, default presets, re-login, and retries at different times of day did not change the Full Rig failure.
Interpretation:
Textures and Joints Only are round-trip bound. From my region the direct route to us-east-1 is unstable, so they fail; a US route fixes them. That points to regional routing or a lack of a nearer endpoint / CDN edge.
Full Rig is a longer server-side solve that exceeds the fixed 300s limit even on a clean US route, so it fails regardless of connection quality, and the timeout cannot be raised client-side.
Net effect: users outside North America and Europe are disproportionately blocked, with a paid VPN as the only partial workaround and no workaround for Full Rig.
Requested resolution (any of these helps):
Investigate AutoRig / texture-synthesis routing for non-US regions, or add regional endpoints / CDN coverage.
Raise the 300s timeout or expose it as a client-configurable setting. The [HTTP] ini values do not affect it, so it needs handling in the plugin.
Investigate the Full Rig job that reaches 75% / RUNNING and then stalls past 300s server-side, also reported on direct US connections.