If you can read a VLESS node name but aren't sure how REALITY differs from Vision, think of them as two separate settings: REALITY handles the outer handshake, while Vision controls how eligible data is forwarded afterward. This guide breaks down both layers using an HTTPS request, shows which fields to check after importing a subscription, and walks through a controlled comparison to tell whether throughput differences come from forwarding overhead.
What VLESS, REALITY, and Vision Do
First, distinguish the HTTPS connection between your browser and the destination website from the connection between your local client and the proxy server. When your browser visits an HTTPS site, it establishes its own TLS session with that site. If the proxy connection also uses a TLS-like security layer, the two layers are nested along the same path. The proxy forwards data sent by the browser; it does not decrypt web pages on the browser's behalf.
- VLESS is a proxy protocol. It carries request details such as user identification and the destination address. In common configurations,
encryption=noneapplies to the VLESS layer; it does not mean the outer connection is unprotected. - REALITY is an outer security mechanism. It handles the handshake between the client and proxy server. The server name used here is not the same field as the address of the website your browser ultimately visits.
- XTLS Vision controls traffic forwarding. In Xray's VLESS configuration, it is specified by
flow. A node using REALITY does not necessarily have Vision enabled too; check the flow value separately.
How REALITY Uses a Website's TLS Handshake
The REALITY server configuration includes a handshake target, and the client sends the specified server name from its configuration. “Uses” here means the outer mechanism draws on the configured target's TLS handshake characteristics and can forward connections that fail REALITY validation to that target. For a proxy session that passes validation, the target site does not authenticate VLESS users or forward your browser traffic on their behalf.
The REALITY client verifies the connected server using the public parameters in its configuration. Once the outer handshake succeeds, VLESS still checks its own user credentials. So being able to open the site mentioned in the configuration in a browser does not mean the node is configured correctly. Likewise, a working node does not guarantee that the proxy can reach every destination website.
- Node address: Determines which machine the client actually connects to and must match the server's listening port.
- Server name: Used in the outer handshake. Use the value provided by the subscription; do not replace it with the address of the website you are visiting.
- Destination website: Comes from the request made by the application. If an error occurs during proxy outbound, check destination reachability and routing rules before changing the server name.
Why Vision Can Improve Sustained Throughput
With ordinary TLS over TLS, TLS records already encrypted by the browser are wrapped in another layer of TLS records. The sender processes the outer records, and the receiver has to unwrap them. This overhead may not be noticeable for a single web request, but during large, sustained transfers, outer-layer encryption and decryption and buffer copying consume more processing time as the data volume grows.
Vision does not skip outer-layer protection from the very first byte. It first identifies the inner traffic; for eligible TLS 1.3 sessions, it can forward subsequent qualifying data directly without applying another layer of TLS record protection to each segment. The original encryption between the browser and website remains in place. Whether this also reduces userspace memory copies depends on the platform, connection type, and forwarding path actually used by the core.
- Differences are easier to measure when: You have a sustained connection, plenty of bandwidth, and a client or server CPU that is already close to its limit. Compare steady-state throughput and CPU usage under the same load.
- Differences may be small when: Requests are brief, the route has packet loss, the exit is rate-limited, or the link capacity is far below what the device can handle. Skipping a layer of record processing cannot reduce the route's round-trip latency.
- Do not generalize this to: Inner plaintext HTTP or sites using QUIC over HTTP/3. The explanation here applies to HTTPS traffic over TCP.
Check the Core and Fields After Importing a Subscription
In v2rayN, double-click the node in the main window to inspect its configuration, then confirm that the node runs on the Xray core. In v2rayNG, open the edit option beside the node to view its details. Field labels may vary between versions, so compare the original field names and values where possible. v2flyNG uses the v2fly core and is not a drop-in alternative for running the same Xray REALITY + Vision configuration.
Protocol and Forwarding
- Core
- Xray
- Protocol
- VLESS
- Transport
- RAW (shown as TCP in some versions)
- Flow
xtls-rprx-vision
These values are for checking the combination discussed here. Do not apply them in bulk to other working nodes.
Outer Handshake
- Security
reality- Server name
- Match the SNI provided by the subscription
- Server public key
- Match the value provided by the subscription
- Short ID
- Keep the value provided by the subscription
If a field is missing after import, first confirm that the core supports it, then reimport the original subscription and check again.
When troubleshooting, do not confuse the local port with the server port. To test the local proxy with another application, first check the current listening settings in v2rayN. If the port has been changed or is in use by another process, copying a commonly used value will only test the wrong endpoint.
Compare Performance on the Same Route
Test Vision's impact on its own only if you control the server and can prepare two matching configurations. Keep VLESS, REALITY, server, and transport the same in both setups, changing only whether Vision is enabled at each end. If you can only choose between different nodes in a subscription, the results can help you pick the better node right now, but they cannot prove that one protocol or flow is faster.
- Use the same request. Use the same device and network to download the same HTTPS file you are authorized to access, such as a 100 MiB test file on your own site. Confirm that both tests use the proxy and HTTPS over TCP, so neither test uses QUIC or bypasses the proxy through a direct-routing rule.
- Record time to first byte and throughput separately. Time to first byte includes the handshake, network route, and origin response; sustained downloads are better for measuring forwarding overhead. Divide the bytes received by the time for the relevant phase, and record CPU usage for the core processes at both ends.
- Alternate the tests. Run each setup five times, alternating between them, then compare the median for each group. Avoid other downloads and make sure the data is not coming from a local cache. If the results overlap widely, retest during a more stable period instead of treating a single peak as a consistent improvement.
Conclusion: Compare Throughput and CPU Usage Together
If throughput is similar in both tests but Vision consistently uses less CPU for the same amount of traffic, count that as reduced processing overhead, not a speed increase. If only time to first byte improves, with no change in steady-state throughput or CPU usage, investigate the handshake and network route rather than crediting Vision.
FAQ: Speed Tests, Handshakes, and Connection Troubleshooting
Does latency dropping from 180 ms to 150 ms mean Vision made the connection faster?
Not necessarily. First check whether the tool measures TCP connection setup, a proxy request, or time to first byte. Retest the same HTTPS address alternately, then compare sustained download speeds separately. These metrics answer different questions.
The handshake fails after import. Which field should I change first?
Do not change anything yet. Reproduce the failure and check the first relevant error in the core log. If the connection is refused or times out, verify the remote address and port. If the connection is established but the handshake fails, check core support, the server name, server public parameters, and the short ID. If only one of several devices fails, also check that device's system clock.
Why can the node's speed test succeed while websites still won't load in my browser?
First confirm that the browser is actually using the proxy. On desktop, check the system proxy or the browser's own proxy settings. On Android, confirm that the VPN is enabled and that the browser is not excluded by per-app rules. Retry while checking the core log for a request to the site. If the request appears, investigate the matched route and any outbound errors.
My old node works reliably. Do I need to switch to REALITY + Vision?
Do not switch just because of the protocol name. Keep the old node and compare both setups over several rounds using sites you visit often and the same download task. If the new node does not improve the throttling or device load you actually experience, keep your current configuration.