For readers who see both REALITY and Vision in a v2rayN node’s details. First, distinguish the proxy handshake from the website handshake. Then see which data can bypass outer encapsulation, and how to assess performance by connection latency, sustained throughput, and core CPU usage.
REALITY Handshake: Using a Site Doesn’t Mean Traffic Passes Through It
When a browser accesses a website over HTTPS, it establishes a TLS session with that site. If traffic goes through a proxy, the proxy client and server can establish a separate protected connection. These connections have different endpoints: one authenticates the website, the other the proxy server. REALITY handles the latter; it does not replace the browser’s verification of the destination website’s certificate.
“Borrowing” here describes the handshake mechanism, not the route that data takes. A REALITY server is configured with a real TLS site and can forward connections that fail REALITY authentication to that destination, where they are handled as ordinary TLS connections. The REALITY server handles proxy connections that pass authentication. The client verifies REALITY using the server public key in its configuration; it does not trust the proxy merely because the destination site presents a regular certificate.
- Connection address: The node address and port determine which proxy server the client connects to first.
- Handshake name:
serverNameis used in the handshake; it is not a domain name to substitute for the node address. - Destination website: The domain the browser requests is determined by the site it accesses. It can be entirely different from the node’s handshake name.
XTLS Vision: Less Encapsulation for Eligible Data
In a typical nested setup, the browser first processes requests and responses into website TLS records. The proxy connection then carries those already-encrypted bytes as its payload. The receiving end must process the outer records before forwarding the inner ones. Besides encryption and decryption, outer-layer processing involves record encapsulation, buffering, and copying data through the software. These costs add up as more data is transferred.
xtls-rprx-vision is a VLESS flow-control mode. With supported transport combinations, Vision checks whether inner traffic meets the conditions for switching—for example, whether it can identify an eligible TLS 1.3 session. The handshake and initial data are still handled through the proxy connection. Once traffic qualifies for direct forwarding, eligible encrypted inner records can be forwarded without wrapping each one in an outer encrypted record. Traffic that cannot be identified or does not meet the conditions remains protected by the outer layer; enabling Vision does not automatically bypass it.
| Processing stage | Standard nested path | Vision path |
|---|---|---|
| Proxy handshake | Establish a protected proxy connection first | A proxy connection and authentication are still required |
| Eligible inner records | Continue wrapping them in outer records | After switching, forward encrypted inner records directly |
| Traffic that does not meet the switching conditions | Handle through the proxy connection | Continue over the protected proxy connection |
Node Settings: Check the Security Layer and Flow Control Separately
REALITY and Vision belong in different fields. REALITY is a transport security layer; Vision is VLESS flow control. After importing a subscription, open the node editor from the main v2rayN list and compare its settings with the parameters provided by your server. These cards illustrate common combinations; they are not interchangeable configuration templates.
VLESS + REALITY + Vision
- Protocol
vless- Transport
- RAW (shown as TCP in some interfaces)
- Security
reality- Flow control
xtls-rprx-vision
Enter the handshake name, public key, and Short ID as specified by the node provider.
VLESS + TLS + Vision
- Protocol
vless- Transport
- RAW (shown as TCP in some interfaces)
- Security
tls- Flow control
xtls-rprx-vision
This combination uses standard TLS certificate verification, not a REALITY public key or Short ID.
Both cards use the same flow-control value but different handshake settings. When copying a node for testing, changing the security-layer dropdown alone will not turn a REALITY node into a standard TLS node. The server must support the selected transport and authentication method. If fields are missing after import, first check the selected core and import format, then verify the original node parameters. Don’t copy identity fields from another node.
Performance Comparison: Test Sustained Throughput and Time to First Byte Separately
The most direct benefit of reducing redundant encryption is lower processing cost per unit of data. Throughput may improve if the device or server is CPU-bound. If the network link is already saturated, the speed reading may barely change, even as core CPU usage falls. Network round trips, queueing, and retransmissions caused by packet loss are unaffected by removing an encryption step on the client.
- Prepare two configurations in an environment you’re authorized to use, with matching server and client settings. Enable Vision in one and disable it in the other. Keep the proxy host, outer security method, core version, and network egress the same; don’t use subscription nodes in different regions as a substitute for a controlled comparison.
- Download the same file from the same HTTPS test endpoint, using the same application transport in both cases—for example, HTTPS over TCP. Alternate between the two configurations so one isn’t always tested when the network is less busy. Make sure the request actually goes through the proxy rather than being served from the browser cache or bypassed by a direct-connection rule.
- Record time to first byte, steady-state throughput, and proxy core CPU usage separately. Use the median for each group and note the range of variation. To compare CPU usage, choose intervals with similar throughput; otherwise, higher usage may simply mean more data was processed in the same amount of time.
Find the Bottleneck: Compare CPU Usage at Similar Throughput
If the two groups deliver similar throughput and the Vision group consistently uses less CPU at the same rate, count that as a reduction in processing cost. If only time to first byte changes, check connection reuse and network round-trip differences. Don’t attribute the change directly to removing outer-layer encryption.
FAQ: Importing, Compatibility, and Speed
My subscription has REALITY. Can I just change Flow to Vision?
Don’t change it on one side only. First check the protocol and original flow-control value in the node details, then ask the provider whether the server is configured to support that flow control for your account. If the server is confirmed to support it, use the complete node configuration provided; don’t replace just one field.
The same node won’t connect after switching devices. What should I check first?
First confirm that the core in use on the device supports this combination. Then compare the node address, port, handshake name, and flow control with those on a device where it works. On desktop, check which core v2rayN is actually using. On Android, don’t assume v2rayNG, which uses the Xray core, and v2flyNG, which uses the v2fly core, are interchangeable. If the logs show a connection timeout first, check network reachability before changing handshake settings.
My speed test is 10–20 ms. Why hasn’t downloading gotten faster?
Check what the test measures: connection time, response time, or file transfer rate. A short request may finish before sustained transfer begins. Test with the same HTTPS file over a sustained download, confirm the actual forwarding path, then compare data transferred per second and core CPU usage.
With Vision enabled, does ordinary HTTP also skip an encryption layer?
Enabling flow control alone does not switch traffic to direct forwarding. Ordinary HTTP has no inner TLS-encrypted records for Vision to forward directly. If the application supports HTTPS, use it and keep website certificate verification enabled; the proxy handshake does not replace transport protection between the application and the website.