Skip to content
Available now

Sylnar CS2

A hardware-driven overlay for Counter-Strike 2. It reads the game from a second machine, works out what you could genuinely see, and draws only that. The aim plans every movement from scratch, and it has good rounds and bad ones.

Adaptive formPlanned per flickNo two alikeShaped speed
Buy now
Beta slots are limited. Pricing locks in now and climbs as each wave fills.
PresentationSylnar CS2, live
The overlay in a live match — visuals, aim and radar as they ship.
250/s
Refresh
3.9 ms
Avg update
99.2%
Uptime
01 / The aim model

No two movements are ever the same.

Most aim assistance travels a curve. A curve is a shape, and a shape on its own is the easy part. What actually separates a person's movement from a machine's is when the speed peaks, how long the whole thing takes, and whether the answer changes from one engagement to the next. Sylnar plans all three, freshly, every single time.

22 FLICKS, ONE TARGET0 REPEATSvertical scale expanded so the shape is visibleS-curve 27%

22 flicks to one target, 22 different paths

The route is planned again from nothing every time. Most of them bow to one side, leaning early or late; the rest trace a shallow S, which is what a hand does when it corrects part way through. Same start, same target, and no two the same. These are not illustrations, they are the real output.

an arm: builds, peaks, eases offa machine: peak on frame oneSPEEDTIME

The speed is the hard part, so the speed is modelled

A machine is fastest on its very first frame and slows from there. An arm builds, peaks, and eases off, and the peak does not sit in the middle. Every movement here is given its own profile, so the moment it is travelling fastest is never twice in the same place.

DURATIONDISTANCEr = 0.86

Further away takes longer, as it should

A person cannot cross a wide angle as quickly as a small one, and how the two relate is not a matter of opinion. Plot what Sylnar plans against how far it has to travel and the points land on that relationship, at r = 0.86. Nothing snaps, nothing is instant, and nothing takes the same time twice.

Four questions, answered separately

How long should it take?
longer the further it goes, never the same twice
What route should it take?
planned fresh, and almost never straight
How is the speed spread?
builds, peaks off centre, eases off
How precise should it be?
less when fast, more when slow, nothing when still

These are four different questions, and a person answers all of them at once without thinking about it. Most products answer one and treat the rest as decoration. Answering the last one properly is what makes the other three fall into place instead of needing to be faked.

Answered together, not chosen between

The Humanizer slider moves the whole distribution rather than just widening it. Turn it down and the movement is clean and direct. Turn it up and it is slower, looser and more deliberate, the way a real one is.

Only in SylnarAdaptive form

It has good rounds and bad rounds, like you do.

Every other layer shapes a single movement. This one carries across the whole match. Sylnar keeps a single number for how well it is going, and that number scales everything else.

DIALLED INquicker to react, tighter, steadier holdRATTLEDslower, looser, likelier to overshootquick killa fight that went nowhereONE ROUNDsimulated with the model's own constants
A kill raises it
And a fast kill raises it more than a drawn-out one. Success genuinely does make the next movement quicker and more direct in people, so it does here.
A fight you lose drops it
An engagement that runs long and ends without a kill knocks it down hard. The next movement is slower, looser and likelier to overshoot. People do exactly this after something goes wrong, and it is one of the most reliably measured things about them.
It recovers, then drifts
Form eases back toward a baseline over a few seconds. The baseline itself wanders across the session, so there is a warm-up early and a slight fade late.
Both hands agree
Recoil control reads the same number. Holding a spray perfectly steady while the aim is visibly rattled would be two hands disagreeing, and no one has those.
Reaction time
scaled by form
Movement duration
scaled by form
Overshoot odds
scaled by form
Steadiness
scaled by form

The swing on any one shot is deliberately small. The point was never a dramatic difference between two flicks; it is that the four hundredth engagement of a session does not look like the third.

02 / Capabilities

Everything in the build.

01

True line of sight

Visibility is ray-cast against the map's real collision geometry. An enemy reads as visible only when there is a clear line to their centre mass, not because the game happened to flag them.

02

Player visuals

Boxes, skeletons, health bars, names, the weapon they are actually holding and distance. Colour follows real line of sight, so what you see reflects what you could see.

03

Motion shaping

Aim movement is shaped rather than snapped: humaniser, speed, precision, pursuit, separate left/right and up/down pull, and falloff at range.

04

Engagement timing

Spot time, minimum and maximum reaction, target-switch cooldown and kill hold. Every stage of an engagement has its own timing rather than one global delay.

05

Situational gates

Rules for when not to act at all: ignore targets behind walls, and hold off entirely while flashed.

06

Recoil control

Compensation with an adjustable variance, driven from live weapon state rather than a recorded spray table, so it follows what the gun is actually doing.

07

Magnetic trigger

Reaction delay range, magnet field of view, pull strength and smoothing, target lead, tap spacing, and an option to act only while scoped.

08

In-overlay radar

A rotating top-down radar with range rings, a view cone and edge clamping, drawn in the overlay itself. No second screen and nothing to keep in sync.

09

Awareness filter

Optionally limits the display to contacts you could plausibly know about (seen, heard by footstep or gunfire, or called by a teammate) with a memory window that fades them out.

10

Auto-scope

Pre-scopes a sniper when a visible enemy comes into range, and drops back out once the angle is clear again.

11

Per-weapon profiles

Save a full setup per gun and have it switch automatically as you change weapons in game, so an AWP and an MP9 never share one compromise.

12

Tuned for the machine

Entity rate, lighter reads and a low-end preset let the same build run well on a modest second PC.

03 / How it behaves

What the software is actually doing.

These are the real behaviours, not marketing shapes. The aim model was developed against an offline simulator driven by recoil sampled from live play, which is how the settling behaviour below was verified rather than guessed at.

CROSSHAIRTARGETinstant snapshaped approach

Aim travels, it does not teleport

A snap would put the crosshair on target in a single frame. Instead the movement is shaped: it builds speed, then eases as it arrives, with the curve governed by the speed, precision and pursuit settings.

FIELD OF VIEWDEADZONEoutside: untouchedinside: assist engagescentred: assist releases

Two boundaries decide everything

Outside the field of view the assist does nothing at all. Inside it, the pull engages. Inside the deadzone at the centre it hands control straight back, so the last part of the movement is always yours. Both radii are yours to set.

on targetERRORTIMEslight overshoot, correctedtarget acquired

A closed loop, so it corrects itself

Error is measured every frame and fed back in, so the aim converges on the target rather than depending on one perfect calculation. A slight overshoot is corrected on the following frames instead of persisting.

NO COMPENSATIONwalks off targetSYLNARstays where you put it

Recoil, cancelled as it happens

Compensation is driven from the weapon's live state rather than a recorded spray table, so it follows what the gun is actually doing on this shot. What remains is the game's own inaccuracy, which nothing can remove, and an adjustable variance so the result is not mechanically perfect.

TIMEcrosshairon a bodyidentityenemy, aliveline of sightnot through a wallreaction delayvaries every shotFIREany check fails→ nothing happens, the trigger is simply not pulled

The trigger asks four questions before it fires

A body under the crosshair is not enough on its own. It has to be an enemy, alive, and genuinely in line of sight, and then a reaction delay has to elapse. That delay is drawn from a range you set rather than being a fixed number, so no two shots share the same timing. If any check fails, nothing happens at all: the trigger is simply not pulled.

GEOMETRYYOUHIDDENVISIBLE

Visibility is proven, not assumed

A ray is cast from your eye to the enemy against the map's real collision geometry. Blocked reads hidden, clear reads visible. Where it cannot prove a line, it says hidden rather than guessing.

GAMING PCruns the gamenothing installedgame stateone way, read onlySECOND PCSylnar runs heredraws the overlay

Two machines, one direction

The gaming PC runs the game and nothing else. Game state is read across to a second machine, which rebuilds the world and draws the overlay on its own display.

04 / What is different

Why this one and not another.

Most of what separates Sylnar is not a longer feature list. It is a set of choices about restraint: what it refuses to do, what it insists on proving first, and how much of the movement it leaves to you.

TYPICALSYLNARruns on the gaming PCruns on a second machinecolours through wallsline of sight is checkedinstant snap to targetshaped, self-correctingfires the moment it canwaits, and the wait variesone profile for everythingsettings per weapon
It proves visibility, it does not assume it
Every visible marker is backed by a ray cast against the map's real collision geometry. Where a line cannot be proven, it reports hidden rather than guessing in your favour.
Timing is a range, never a constant
Reaction, target switching and the hold after a kill each have their own window. Nothing in the engagement happens on a fixed interval.
The last part of the aim is yours
The deadzone exists so the assist lets go before the shot. It closes the gap; you take it from there.
One setup per weapon, switched for you
An AWP and an MP9 want opposite things from every slider. Profiles follow the gun in your hands instead of forcing one compromise across the whole round.
Built to run on a modest second machine
Read rates, lighter entity passes and a low-end preset came out of making it work on cheap hardware, not out of a spec sheet.
05 / Interface

Six tabs, every setting explained.

The overlay opens with a single key. Every control carries a plain-language explanation of what it does when you turn it up and when you turn it down.

Sylnar CS2 boot splash
The loader, on launch.
Aimbot
Sylnar CS2, Aimbot tab
Targeting, motion shaping, engagement timing, situational gates and recoil control.
Trigger
Sylnar CS2, Trigger tab
Reaction timing, the magnetic pull controls, and the conditions that must hold before it fires.
ESP
Sylnar CS2, ESP tab
Which visuals are drawn, the distance filter, and the colours for visible, hidden and team.
Misc
Sylnar CS2, Misc tab
Radar layout, the awareness filter, performance controls and auto-scope.
Config
Sylnar CS2, Config tab
Mouse calibration, named configurations, per-weapon profiles and the input device.
06 / How it runs

Two machines, one link.

01

Gaming PC

Runs the game exactly as it always did. Nothing is installed on it and nothing about how you play changes.

02

The link

A capture card reads memory over PCIe and hands it to the second machine, hundreds of times a second.

03

Second PC

Sylnar runs here: it rebuilds the world, ray-casts visibility against the map, and draws the overlay on its own display.

What you need

  • A second Windows PC with a spare PCIe slot
  • A supported DMA capture card
  • A Makcu input device
  • A monitor or capture path for the second machine

SYLNAR CS2

Accounts are open. Create one now and your licence, downloads and per-weapon profiles all live in the same place, including for whatever we build next.