VLESS handles proxy requests, user identification, and data; REALITY secures the proxy handshake and authenticates the connection; XTLS Vision can reduce the overhead of re-encapsulating eligible inner TLS traffic. This guide is for users who have already imported a node and want to understand its settings. You’ll learn to distinguish the node address, handshake domain, and website domain—and whether to investigate the connection, authentication, or forwarding path.
HTTPS Through a Proxy: What Each TLS Layer Does
When you visit an HTTPS site in a browser, the browser still establishes a TLS session with the site and verifies its certificate. Separately, the proxy client connects to the proxy server. With conventional TLS-in-TLS forwarding, the browser’s already encrypted data is wrapped in outer encrypted records between the client and proxy server. The two layers protect different endpoints, so a successful outer handshake does not mean the website is reachable.
- VLESS: Carries the user identity, destination address, and data needed for proxying. In the setup discussed here, the selected security layer provides outer protection; the VLESS protocol itself does not determine it.
- REALITY: Secures the handshake between the client and proxy server. It does not verify the website the browser ultimately visits or determine which domain the browser opens.
- XTLS Vision: Enabled through VLESS’s
flowoption. With compatible security and transport settings, it changes how some data is forwarded; it is not a separate handshake protocol that replaces TLS.
Where XTLS Vision Saves Overhead: The Data Phase
Vision inspects inner traffic. Once both ends confirm that direct forwarding is possible, eligible inner TLS records can be forwarded without wrapping each one in outer encryption. Browser-to-website encryption remains in place. Direct forwarding does not start with the first byte, and ordinary plaintext traffic cannot use this path just because Vision is enabled. A node name alone cannot tell you whether a particular connection used direct forwarding.
For a TLS 1.3 application-data record using AES-GCM, add 1 byte for the inner content type as well as the record header and authentication tag. The 22-byte figure above excludes padding, TCP/IP headers, and handshake traffic. It is a calculation based on the record format, not a measured result or a guaranteed reduction in every network packet. Avoided overhead may also include outer-record encryption and decryption, but this does not imply a fixed throughput gain.
How to Test: Measure Large Transfers and Time to First Byte Separately
Using the same device, server, and download target, compare two compatible configurations with all other settings unchanged. Record sustained throughput during a large-file transfer, CPU usage by the core process, and time to first byte on a new connection. If CPU usage drops during downloads while time to first byte stays about the same, that does not mean Vision had no effect: the metrics reflect different stages.
REALITY Handshake: Where the Real-Site Certificate Appears
“Using a real-site certificate” is shorthand for how the handshake appears externally; the proxy server does not contain the real site’s certificate and private key. For a proxy connection authenticated by REALITY, the client verifies the handshake proof using the configured server public key, while the server uses the matching private key. For an ordinary TLS connection that fails REALITY authentication and is forwarded to the destination site, the real site presents its certificate during its own TLS handshake.
- The node address and port determine where to connect. The client first connects to the proxy address provided by the subscription.
443is a common HTTPS port, but it is not required by REALITY. Use the port specified in your subscription. - The handshake domain determines which server name is sent.
serverName(often shown as SNI in clients) must match a name allowed by the server. It does not have to match the node address, and it does not change when the browser opens a different website. - The public key and Short ID must match the server. The public key corresponds to the server’s REALITY private key; Short ID is a REALITY matching field. They are separate from the VLESS user UUID and must not be interchanged.
- Check the outbound route after authentication. Once the handshake is established, the proxy server still has to handle the destination connection requested by VLESS. If the handshake succeeds but a particular HTTPS site does not load, check routing, outbound DNS resolution, and the connection to the destination site.
v2rayN and v2rayNG: Check VLESS, REALITY, and Flow
In v2rayN, double-click the relevant entry in the node list to open its editor and check the fields. In v2rayNG, tap the pencil icon for the entry in the configuration list. Field names can vary between client versions, especially for transport type and flow control. If the subscription already provides values, check those first; don’t change the node’s settings just to match an example.
VLESS Proxy and Transport Settings
- Protocol
- VLESS
- Address and port
- Use the node address and port from the subscription
- User ID
- UUID assigned to this node
- Transport type
- Often TCP or RAW; use the client interface and node settings as the source of truth
- Flow control
xtls-rprx-vision; use only if required by the server
Flow control must match the user configuration on the server. Changing the security type does not automatically fix a flow mismatch.
REALITY Handshake Settings
- Security type
reality- SNI
- Node-specified
serverName - Public key
- Server’s corresponding public key, usually shown as
pbkin a share link - Short ID
- Value specified by the node, usually shown as
sidin a share link - Fingerprint
- Client handshake fingerprint specified by the node, such as
chrome
The handshake fingerprint affects how the client initiates the handshake. It is not the website’s certificate fingerprint and does not replace the public key.
When using these settings in v2rayN on desktop, check that the active core is Xray and supports this combination. On Android, check the relevant entry in v2rayNG with the Xray core. v2flyNG uses the v2fly core; recognizing a link prefix does not by itself mean it supports the same flow-control and security combination.
REALITY Won’t Connect or Shows No Speed Gain? Troubleshoot by Connection Stage
Changing the SNI caused a timeout. What should I check?
Restore the SNI from the subscription and test the same node again. If the core log reports a timeout connecting to the node address and port, first check whether that address and port are reachable. If the connection is established but the handshake fails, check the SNI, public key, and Short ID one by one. Being able to open a domain in a browser does not prove the proxy server allows it as an SNI.
Vision Is Missing After Import. Should I Enter the Flow Manually?
Open the node editor and confirm that the protocol is VLESS, then check “Flow” or “Flow control.” If you can view the original share link, check whether it includes flow=xtls-rprx-vision. Don’t add this value if it wasn’t in the original configuration. If it was present but isn’t shown in the client, first check whether the import completed correctly and whether the active core is compatible.
Both Nodes Work. Why Isn’t Vision Faster?
Latency tests mainly reflect network round-trip time, handshake time, and the response time of the test target. First confirm that the test request actually uses that node, then download the same large HTTPS file from the same target repeatedly and compare sustained performance. If the connection is already bandwidth-limited, reducing encryption and decryption work may not increase download speed.
Node Test Passes, but Websites Won’t Load in the Browser. What Should I Check First?
First confirm that the browser is actually using the client. In system proxy mode, make sure the browser follows the system proxy settings and that its proxy configuration matches the client’s local listening port. If the log already records the browser request, check which route matched it and whether the selected outbound reported an error. A successful node test only confirms that the test request worked; the browser request may take a different path.