For readers who have imported a subscription into v2rayN and want to understand what REALITY and Vision each do. First, distinguish the proxy handshake from the website's TLS handshake. Then see which traffic may avoid an extra layer of encryption, and use node settings and speed tests to pinpoint issues.
The REALITY handshake: Proxy identity and website certificates are verified separately
When you open an HTTPS page, your browser verifies the website's identity. With a proxy enabled, the Xray client must also establish a connection to the proxy server first. These are two separate exchanges: the first protects the browser-to-website session; the second creates a tunnel between the client and proxy server. REALITY is involved in the latter.
The diagram shows a successful new connection. The Xray client first starts a TLS 1.3 handshake and uses preconfigured public-key-related parameters to authenticate the REALITY server. Once the handshake succeeds, the client sends a VLESS request through the tunnel. For HTTPS traffic, the browser then checks the destination site's certificate in its own TLS session.
- “Certificate borrowing” is an informal description. The REALITY handshake can refer to a target site configured on the server. Normal proxy authentication does not require the target website's private key, and a publicly available certificate alone cannot be used to impersonate a website. The target site's genuine certificate does not, by itself, authenticate the proxy.
- The node address and SNI serve different purposes. The address and port determine which proxy host the client connects to; the node's SNI is the name used in the proxy handshake. When you visit a webpage, the browser handles the page's own domain. The two do not have to match.
- A successful handshake does not mean the website is reachable. DNS resolution for the target domain, routing rules, or the server's outbound connection can still fail after proxy authentication succeeds. If the connection is established, don't immediately re-enter the handshake settings; first check which outbound route the request actually uses.
XTLS Vision: Which layer of encryption does it reduce?
In a typical nested setup, the browser encrypts webpage data into TLS records, then the proxy tunnel encrypts those records again as payload. The receiving end has to remove the outer layer first. This extra processing happens along the proxy path; it does not mean the browser sends the same webpage request twice.
Here, Flow refers to how VLESS handles a data stream, not a rate limit. Early in a connection, Vision processes and pads data according to its rules, while identifying inner TLS traffic. When both sides support it and the conditions for switching are met, eligible records can be forwarded directly without continuously wrapping them in the proxy tunnel's TLS record layer. The webpage's own TLS encryption remains intact; this path does not apply to ordinary plaintext traffic.
| Stage | What happens | Performance implications |
|---|---|---|
| Proxy handshake | Establishes the tunnel and authenticates the proxy | Handshake latency remains; direct forwarding does not eliminate it |
| Eligible inner TLS data | After switching, forwards the existing encrypted records | Differences in outer-layer processing are more likely to show up with high traffic volumes |
| Data that does not meet the switching conditions | Continues through the protected proxy path | Throughput cannot be estimated as if it were directly forwarded |
So REALITY is not another name for Vision. REALITY handles the proxy tunnel's handshake and security; Vision determines how eligible data travels through that tunnel. Vision mainly reduces the encryption, decryption, and record-processing overhead for each chunk of data. It cannot shorten the physical route or eliminate retransmissions caused by packet loss.
v2rayN settings: Check transport, security, and Flow
In the v2rayN main window, double-click an imported node to open its settings. First, confirm that the node uses VLESS and that the active Xray core supports this combination. Then check the settings below. Labels can vary between versions—for example, bare TCP may appear as tcp or raw in the interface or configuration. There's no need to rebuild a working node just to standardize the label.
Connection and Flow
- Protocol
VLESS- Address and port
- Use the proxy endpoint provided by your subscription
- User ID
- Must match the server configuration
- Transport
- Check against the node configuration; a typical setup uses bare TCP
- Flow
xtls-rprx-vision; must match the server-side user settings
After importing a subscription, check the original values first. Adding Flow to an existing VLESS node on its own does not enable Vision on the server.
REALITY handshake settings
- Security type
reality- SNI
- Handshake name allowed by the server
- Public-key-related field
- Use the client value provided by your subscription
- Short ID
- Must match a value allowed by the server
- Fingerprint
- Keep the client handshake fingerprint specified by the subscription
SNI is not a speed-test address. Never enter the server's private key on the client, and don't include your full subscription link in troubleshooting screenshots.
Keep the imported configuration unchanged, then test with a browser known to use the system proxy. If the browser works but a particular desktop app doesn't, check whether the app uses the system proxy or needs a proxy configured separately. Don't switch TUN and change node settings in the same test; otherwise, it's hard to tell whether traffic routing changed or the node started working.
Speed tests: Separate first-byte time, throughput, and CPU usage
A single latency reading can't tell you whether Vision reduced processing overhead. Latency is more affected by the handshake, network route, and remote response; a sustained download is a better way to reveal throughput limits. The figures below are repeatable self-test settings, not protocol benchmarks or v2rayN defaults.
- Use a service you're authorized to test, and prepare two configurations that both connect successfully. Keep the proxy server, test source, and local network fixed. Use the same large HTTPS file to avoid mistaking a cache read for a network download.
- Alternate between the two configurations. In each round, record time to first byte, sustained throughput, and CPU usage by the core process. Compare medians, not peak results. If the configurations require different servers or routes, you can compare real-world experience, but you can't attribute the difference to Vision.
- If time to first byte is similar while sustained throughput increases and CPU usage drops at the same rate, the results are consistent with reduced processing overhead. If throughput varies by time of day and CPU is far from saturated, check outbound bandwidth, remote rate limits, and background downloads on your PC first.
Takeaway: Find the throughput bottleneck first
Check the core process and network speed in Task Manager at the same time. If network throughput is already close to your access limit and the core's CPU usage is not under significant pressure, record the network baseline first. Switching Flow is unlikely to produce a higher speed-test result.
Common issues: No connection after import or no speed improvement
Why is Flow blank when my subscription includes REALITY?
They're independent settings. First check whether the subscription explicitly specifies Vision; if it doesn't, leave Flow blank. If the original configuration does specify Flow but it isn't shown in the editor, check the v2rayN and Xray core versions in use, then reimport and check again. Don't overwrite an existing working node.
The handshake fails when I select the node. What should I check first?
Start with the first connection error in the main-window log. If the TCP endpoint times out, check the address and port. If the endpoint connects but the handshake fails, compare the SNI, public-key-related values, and Short ID with the subscription. Also check that Windows time is synchronized under Settings → Time & language → Date & time.
Why is my download no faster even though webpages load?
First confirm the download is actually going through that node and isn't bypassing the proxy because of a routing rule. Then run a sustained test using the same download source. Short requests are often dominated by round-trip latency. If bandwidth is already saturated, download speeds may remain unchanged even when the core has less processing to do.
Do I need TUN enabled to use Vision?
No. TUN and system proxy settings determine how app traffic reaches the client; Vision determines how eligible traffic travels over the proxy connection. Test the node using the system proxy first. Only test TUN separately if an app doesn't use the system proxy and you need transparent traffic routing.