REALITY and XTLS Vision Explained: Handshake Masking, Flow Control, and Performance

Learn how REALITY disguises its handshake and how XTLS Vision reduces redundant encryption in Xray. Understand how the two work together.

At a glance

For users who have imported a node and want to understand its security type and Flow setting. REALITY handles the outer handshake; Vision changes how eligible data is forwarded. Knowing the difference helps you tell handshake failures from speed changes worth retesting.

REALITY Handshake: Two Separate Connections

App requestProxy connectionOuter handshakeProxy forwardingDestination traffic

When your browser accesses an HTTPS site through a proxy, it still has its own TLS session with the site. The client core establishes a separate connection to the proxy server. REALITY handles the security handshake on this second connection. The diagram groups steps by function; it is not a packet-capture timeline. The proxy connection may be established in advance, and the browser may reuse an existing connection to the site.

Handshake masking means REALITY uses a TLS-compatible handshake pattern and handles unauthenticated connections differently. A normal proxy session requires preconfigured authentication material; simply being able to reach a particular website does not grant access. The server name used in the handshake is also distinct from the destination address requested by the app.

XTLS Vision: Less Redundant Processing, Not a Replacement for HTTPS

With ordinary nested encryption, the proxy's outer encrypted records carry the TLS ciphertext generated by the browser. Both ends of the proxy must also encrypt and decrypt this outer layer. For connections carrying large amounts of HTTPS data, that extra work can use significant CPU. With small transfers or a slow network, however, it may not be the main bottleneck.

Vision is a flow-control mode used with VLESS in Xray. It identifies inner TLS traffic and handles padding and encapsulation at the start of a connection. For eligible encrypted data, it can switch to direct forwarding in the relevant direction, reducing redundant outer-layer processing. This does not mean the entire connection is passed through directly from the first byte, nor is Vision a new encryption protocol between the browser and the website.

Outer handshake

Node security type
reality
Scope
Client core to proxy server
Primary purpose
Authentication and secure handshake

A security type of REALITY does not mean Vision is enabled too.

Data flow control

Proxy protocol
VLESS
Flow
xtls-rprx-vision
Source of performance gains
Less redundant outer-layer encryption and decryption for eligible data

Flow is a configuration agreement between both ends of a node, not a general-purpose local speed toggle.

If a node name includes “Vision,” check the actual settings first. Subscription labels are set by the provider; the core runs the parsed configuration. Also, direct forwarding is not the same as system-level zero-copy: whether a lower-overhead data path is used depends on the operating system, connection type, and core implementation.

Speed Tests: Measure Time to First Byte and Sustained Throughput Separately

Time to first byte depends on connection setup, network round trips, and the destination's response. Sustained throughput is more likely to reflect encryption and data-transfer overhead. If the two routes use different exits, return paths, or destination servers, a single download test cannot attribute the difference to Vision. The figures below are test parameters, not measured results.

443
Default HTTPS destination port
100 MiB
Example test file for sustained throughput
5 runs
Recommended runs per group
  1. Use two configurations that have already been verified to work. Keep the server, exit route, destination, and transport conditions the same where possible. Do not change Flow only on the client and treat a failed connection as a valid comparison.
  2. Keep the access method consistent. Use the same test file and avoid hitting a local cache; for cold-connection time-to-first-byte tests, start a new connection each time. Record long downloads separately from short requests so connection reuse does not get mistaken for a flow-control benefit.
  3. Alternate between the two groups and record time to first byte, completion time, and CPU usage of the core process. Compare medians first, then look at the range of variation. If a one-off difference falls within each group's normal variation, avoid drawing conclusions.

How to Interpret Results: Check CPU Before Throughput

If throughput is similar but CPU usage drops, the gain is lower processing overhead, not more bandwidth. If CPU usage is already low and both groups hit the same speed, check for route limits and destination response time first.

Diagnose Node Failures by Stage

Reproduce one clear failure first. Note the node name and time, then request an HTTPS address that usually works. Check only the logs for that request. Avoid switching clients, changing routing, and updating subscriptions all at once; otherwise, it is hard to tell which change affected the result.

Why does the speed test show a result if webpages still won't load?

First check what the test measures: port connectivity, an HTTP request, or a complete download. If it only tests the port, check whether “System proxy” is configured as expected in the v2rayN tray menu, or whether v2rayNG has established a VPN and the app isn't excluded by per-app rules. The test and your browser may not use the same route.

The handshake fails after updating the subscription. Should I change Flow?

First check whether the failure occurs during the outer handshake. Verify the system clock and current Xray core version, and ask the provider whether the node parameters have changed. When comparing old and new configurations, focus on the security type and node address. Never post a full subscription URL or authentication material in public logs.

Will the same node work in every Android client?

Check the core first, then verify the imported settings. v2rayNG uses the Xray core, while v2flyNG uses the v2fly core. Successfully importing a node name does not prove that v2flyNG supports Xray's REALITY and Vision combination. Open the node details after importing to confirm the security type and Flow were not ignored, then test an actual webpage request.

The site reports a certificate error. Does that mean the node handshake failed?

Check the domain in the browser error first, then review the core log from the same time. If the core established a proxy session but the browser reports a certificate error for the destination, check the system clock, the domain you actually visited, and any certificate inspection policy on the network. Do not disable browser certificate verification.

Download V2Ray Clients