YouTube Keeps Buffering on Clash? Fix Playback and Connection Issues
Identify whether YouTube, the browser, or Clash is failing
YouTube that will not load, starts slowly, or buffers repeatedly does not point to one specific cause. The connection may be reaching a congested or unavailable proxy node, a rule may send YouTube traffic through the wrong policy, DNS may return an unusable address, or the browser may be trying to use a transport that the active proxy path does not handle. TUN permissions, another VPN, and browser-specific proxy settings can also change where traffic goes.
Start with a controlled comparison instead of changing several settings at once. Check whether other websites load through the same browser, and whether YouTube works on another device or network. If all websites fail, investigate the client, node, or local connection first. If only YouTube fails, focus on the routing rules, destination domains, DNS, and browser transport. If the YouTube page opens but video playback buffers, the page and video stream may be taking different routes.
In the Clash client, open the connection list or traffic log while loading a YouTube page and starting a video. Look for requests involving domains such as youtube.com, googlevideo.com, ytimg.com, or googleusercontent.com. The exact set varies by page, account, browser, and video. Check the matched rule and selected proxy group for several requests rather than relying on a single entry. The page can load while video segments are routed elsewhere.
If no new YouTube-related connection appears, the browser may not be using this Clash instance. Check for a browser proxy extension, a separate VPN, another proxy client, or a manually configured browser proxy. Also close and reopen the video after changing settings: a browser can keep existing connections alive, so a new rule or node may not affect a stream that is already playing.
Keep a simple record of what you changed and what happened. For example, note the active node, whether the client is in Rule or Global mode, whether system proxy or TUN is enabled, and whether the video begins after switching one setting. This makes it easier to reverse a test and prevents a temporary workaround from becoming an unexplained permanent configuration change.
Test the node and proxy mode before editing YAML
A node can report a low latency while still performing poorly for video. A latency test typically checks whether a small request can reach a test endpoint; it does not measure sustained throughput, packet loss, congestion at peak hours, or access to every YouTube delivery server. If buffering occurs, select another node in the relevant proxy group and test the same video again. Prefer a stable connection over a node that only has the lowest displayed delay.
Check which group actually handles YouTube traffic. A configuration may contain a general group, a media group, a region selector, and automatic latency-testing groups. Selecting a node in one group has no effect if the rule sends YouTube requests to a different group. In the connection details, follow the matched rule to its policy group, then make the test selection in that group. If the interface offers a per-domain connection view, confirm that new video connections use the selected node.
Use Rule mode for the first test unless you have a specific reason to use another mode. In Rule mode, each connection follows the rules in the active configuration. Global mode sends most proxied traffic through the selected global policy, but direct, reject, and special handling can still depend on the client and core. Direct mode generally avoids proxy routing and is not a useful test for a destination that is unreachable on the local network. Do not treat a mode switch as a permanent fix until you have confirmed what it changes.
Check whether the rule set has a YouTube-specific entry or a broader rule that catches its traffic first. Rules are evaluated in order, so an earlier domain, process, or rule-provider match can take precedence over a later YouTube entry. An illustrative set of domain rules might look like this:
rules:
- DOMAIN-SUFFIX,youtube.com,Media
- DOMAIN-SUFFIX,googlevideo.com,Media
- DOMAIN-SUFFIX,ytimg.com,Media
- MATCH,Default
Media and Default above are example policy-group names, not built-in values. Each must match a group defined in the same active configuration. Some configurations already provide media rules or a maintained rule provider; adding duplicate entries may make troubleshooting harder. Before editing a subscription-generated YAML file, check whether the client or subscription update will overwrite direct edits. Use the client’s supported override or rule-editing feature when available.
To distinguish routing from node quality, compare one video under controlled conditions: keep the browser, video quality, and client mode unchanged; switch only the node in the group that handled the request; then check the connection log and playback. If one node consistently works and another does not, the likely issue is the node or its route to the video delivery network. If all nodes behave alike, continue with rule, DNS, and transport checks.
Work through node, routing, DNS, and TUN checks
Use the following sequence when YouTube still buffers. Make one change per test, start a fresh playback session, and record the result. The sequence checks the most common path differences without requiring an immediate client reinstall.
- Confirm the active configuration. Verify that the intended profile is running and that its proxy groups and rules are loaded. A profile name in the sidebar does not always tell you whether an override, merged configuration, or subscription update has changed the runtime rules.
- Check the selected policy and node. In the connection list, identify the policy group used by YouTube video requests. Select another node in that group, wait for existing streams to close, and retry. If the new connection log still names the old node, you may have changed a different group.
- Compare Rule and Global modes briefly. If the client provides a Global policy selector, choose a known working node and test a new playback session. A difference between Rule and Global mode points toward rule ordering or group selection; no difference makes a simple routing mismatch less likely. Restore the normal mode after the comparison.
- Inspect DNS behavior. Check whether the connection log shows the expected domains and whether DNS requests are failing or timing out. If the client has a DNS section, confirm that DNS is enabled and that its upstream servers are reachable through the current network. Settings such as
nameserver,proxy-server-nameserver, andfallbackhave different roles in supported Mihomo configurations; do not copy a DNS block from another profile without checking the core version and the client’s effective configuration. - Clear only relevant stale state. After changing DNS or routing, close the video tab and start a new session. If the client exposes a DNS cache or connection reset control, use it once and retest. Avoid deleting every cache or subscription file as a first step: doing so can remove useful settings without proving that stale data caused the issue.
- Test the proxy path without TUN. If the operating system system proxy is enabled, temporarily turn TUN off and test again. This helps determine whether the problem is related to traffic capture, routing permissions, or a conflict between proxy mechanisms. Restore the original setting after testing.
- Test TUN separately when needed. If the browser or another application is not following the system proxy, TUN may be needed to capture traffic that ignores it. Confirm that the client has the required permissions, the TUN interface starts successfully, and the runtime log does not report a device, driver, or route error. Do not enable multiple VPN or TUN clients at the same time.
DNS and routing interact. A domain rule may be easy to apply when the request still carries a domain name, while an IP-only connection may need DNS mapping, sniffing, or a suitable rule strategy to identify its destination. Mihomo features and available options vary by core version and client. Use the client’s runtime log and the configuration accepted by its installed core, rather than assuming a setting from an older Clash guide exists in every build.
Video playback can also use QUIC, which runs over UDP, commonly on port 443. A browser using HTTP/3 may behave differently from one using a TCP-based HTTPS connection, especially if the current proxy path or node does not handle that UDP traffic as expected. If your browser offers a supported way to disable HTTP/3 temporarily, or you can test in a browser that negotiates a different transport, use that only as a diagnostic comparison. A change in playback points toward a UDP or transport-path issue; it does not by itself prove that QUIC should remain disabled. TUN support, UDP forwarding, and rules can differ by core, client, and operating system.
| Observation | Most useful next check |
|---|---|
| All websites fail through Clash | Check the active node, core status, local connection, and client logs. |
| YouTube page loads but video buffers | Inspect the matched policy for googlevideo.com connections and compare nodes. |
| Rule mode fails but a controlled Global test works | Review rule order, rule providers, and the group named by the matching rule. |
| Playback changes when TUN is disabled | Check TUN permissions, route conflicts, and whether another VPN is active. |
| Playback changes when testing another browser transport | Investigate UDP/443 handling, TUN capture, and the node’s UDP support. |
When testing on Android, also confirm that the Clash-based app is the active VPN service; Android normally allows only one VPN service to hold the VPN slot at a time. On Windows and macOS, check whether a security product, another VPN, or a manually configured system proxy is intercepting traffic. These platform checks matter because a valid configuration cannot route traffic that never reaches its active client.
Know when to update, reset, or reinstall the client
Reinstalling is rarely the best first response to buffering. A client reinstall will not fix a congested node, a rule that selects the wrong group, unreachable DNS upstreams, or a provider-side problem. Before considering it, test a second node, confirm the active policy in the connection log, compare another browser, and check whether the same configuration behaves differently with TUN off. These tests identify whether the issue belongs to the client, core, route, or network.
Update the client or Mihomo core when the installed version is known to lack a required feature, fails to start TUN, or reports a reproducible core error. Check the client’s own release information and its supported core versions; similarly named clients can bundle different kernels. After an update, confirm the core version in the About page or startup log and verify that the active configuration still loads. Do not assume that updating the GUI also updates a separately managed core.
If you reset settings, first export or copy the profile, subscription URL, custom rules, group choices, and any supported overrides. A full reset can remove local changes that are not stored in the subscription. Test with a minimal, known-good profile only if you can do so safely and have a way to restore your existing setup. Avoid sharing subscription URLs, authentication tokens, or unredacted logs when asking for help; they may contain private access information.
A useful stopping point is when the connection log shows YouTube requests using the intended policy and node, DNS succeeds, and a fresh session plays normally. If playback still fails across multiple working nodes and devices, check whether the service provider or network has a temporary issue. If only one device is affected, compare its proxy and VPN state with a device on the same network before changing the entire configuration.
Frequently asked questions
Why does the YouTube page load while the video keeps buffering?
The page and video segments can use different destination domains and may be matched by different rules. Check the connection log while the video is actively playing, especially for googlevideo.com, and confirm which group and node handle those requests. Also consider node throughput, UDP/443 behavior, and DNS rather than assuming the page’s successful load proves the video route works.
Should Clash be in Rule mode or Global mode for YouTube?
Rule mode is usually the appropriate everyday setting when the configuration has working domain and policy rules. A brief Global-mode comparison can help isolate a rule or group-selection problem, but it is not a universal fix and may route unrelated traffic differently. Return to the intended mode after the test.
Should I replace the DNS settings whenever YouTube buffers?
No. First establish that DNS requests are failing or producing a relevant routing problem. DNS settings can depend on the core version, proxy topology, and client overrides. Changing several upstreams at once can create new failures and obscure the original cause. Check the runtime log and make one supported change at a time.
Will reinstalling Clash fix repeated buffering?
Only if the client installation or local runtime is actually damaged. Reinstalling does not improve node capacity or correct an unsuitable rule, DNS path, or UDP route. Test another node and inspect the connections first; back up the profile and custom settings before resetting or reinstalling.