Mouse feels accelerated, floaty or too slow after switching from X11 to Hyprland
Re-checked on this workstation (omarchy 4.0.2-1, omarchy-settings 4.0.2-1, hyprland 0.56.2-1, kernel 7.1.9) against the current Hyprland wiki, libinput's own documentation and Hyprland's source at tag v0.56.2. Every version-sensitive claim in the record held. The wiki config-options table still says sensitivity is float, default 0.0, limits -1.0 to 1.0. accel_profile is string, default [[Empty]], "Leave empty to use libinput's default mode for your input device", options adaptive, flat and custom. force_no_accel is bool, default false, with the wiki's own "not recommended due to potential cursor desynchronization" note. The custom form custom <step> <points...> with the example custom 200 0.0 0.5 is verbatim from that page. The devices page still gives hl.device({ name = ..., ... }) and still excludes force_no_accel and the window-management options from a device block, and I confirmed that in source: src/managers/input/InputManager.cpp reads force_no_accel only as the global input:force_no_accel, while sensitivity and accel_profile go through getDeviceFloat and getDeviceString with a per-device bare key. libinput's pointer-acceleration page confirms the cause in its own words: "The adaptive profile is the default profile for all devices" and "The flat profile is simply a constant factor applied to all device deltas, regardless of the speed of motion". One thing the previous auditor could not check is now confirmed in source rather than inferred: the record's touchpad device block sets bare natural_scroll and scroll_factor rather than nesting them under touchpad = { ... }, and that is correct. InputManager.cpp resolves a per-device key by its bare name and chooses the fallback by device type, so natural_scroll in a touchpad's block falls back to input:touchpad:natural_scroll and in a mouse's block to input:natural_scroll. Confirmed live on this machine: hyprctl getoption input:sensitivity returns float: 0.000000 with set: true, and hyprctl getoption input:accel_profile returns str: [[EMPTY]] with set: false, which is the record's central claim about Omarchy proved on hardware. Four things needed correcting. First, the device names were invented. logitech-g502-hero-gaming-mouse is not a name anything printed, and the record gave no rule for the lowercase-hyphen form, so the corrected fix prints them with hyprctl -j devices and uses real ones from this machine: razer-razer-basilisk-v3 and corsair-corsair-gaming-k55-rgb-keyboard-1. Second, the Omarchy path was right but the instruction was not. ~/.config/hypr/input.lua here is byte-identical to the packaged template at omacom/omarchy@quattro config/hypr/input.lua and ships fully commented out, and lines 22 to 26 of that template already carry sensitivity = 0.35 and accel_profile = "flat". The Omarchy fix is therefore uncommenting two lines, not writing a new block, and a reader who uncomments the whole block instead picks up kb_layout = "us,dk,eu" and a kb_options string that replaces rather than merges and so drops Omarchy 4's compose:caps,shift:both_capslock_cancel. Third, the record's step 1 told the reader to set sensitivity = 0.0, which is already what /usr/share/omarchy/default/hypr/input.lua sets, so on Omarchy that line changes nothing and reads as a failed edit. Fourth, and this is the one the record was most incomplete on, it said flat means no acceleration without ever saying what flat costs. The corrected fix gives libinput's own description and the consequence: a slow movement and a fast flick use the same factor, crossing a wide desktop takes more physical travel, raising DPI is the usual answer, and adaptive is normally better on a touchpad. The record also never told the reader how to confirm the change took, so a step 7 now gives hyprctl configerrors and the getoption before-and-after values measured here, with the caveat that getoption is global only and a per-device override cannot be read back that way. Severity and frequency held at medium and very-common, and the cause was correct on every point, so it is reissued only to add the Omarchy facts and the cost of flat. Sources: the two raw wiki URLs the record cited have moved in the hyprland-wiki tree, so I retrieved the current paths (content/configuring/core/config-options.md and content/configuring/core/devices.md) and added them, along with libinput's pointer-acceleration page and the quattro-branch Omarchy files. Removed the raw.githubusercontent.com/basecamp/omarchy/dev/config/hypr/input.lua URL: it still 200s through the rename redirect, but the repo is omacom/omarchy now and dev is not the branch Omarchy 4 ships, so the quattro URL replaces it. Kept hyprwm/Hyprland#7088 with lower weight: it resolves and it does show a user whose global sensitivity did nothing and whose per-device block worked, but it is from July 2024 and v0.41.2, predates the Lua config, and was closed in a bulk move to discussions rather than resolved. It is a symptom report, not evidence for 0.56 behaviour. NOT exercised: nothing was written, and no reload, eval, keyword or dispatch was run, because this is the operator's live graphical session with the mouse in use. There is no touchpad on this machine (hyprctl -j devices reports an empty touch list), so every touchpad claim comes from the wiki, from Omarchy's packaged input.lua and from InputManager.cpp, not from hardware. Neither accel_profile = "flat" nor the custom curve was applied here, so their feel is reported from the documentation only.
force_no_accel is documented as not recommended because it can desynchronise the cursor from real pointer position. Prefer accel_profile = "flat".
Symptom
"My mouse feels completely wrong under Hyprland." The pointer accelerates when you move fast and crawls when you move slow, so aiming in games and precise clicking are impossible. Or it is simply far too slow/too fast compared to the same mouse under X11 or Windows. Setting input { sensitivity = ... } seems to do nothing above a certain value, and turning the sensitivity up to fix the mouse makes the touchpad unusably twitchy at the same time.
Cause
libinput applies its adaptive pointer-acceleration profile by default, and its own documentation calls adaptive "the default profile for all devices". That is the same curve X11 users usually disabled long ago with libinput Accel Profile Flat in an xorg.conf.d snippet, which no longer applies on Wayland. Hyprland ships no acceleration override of its own: input.accel_profile defaults to empty, meaning "use libinput's default mode for your input device". Omarchy 4 does not override it either. /usr/share/omarchy/default/hypr/input.lua sets sensitivity = 0 and says nothing about accel_profile, confirmed on a running install where hyprctl getoption input:accel_profile reports str: [[EMPTY]] with set: false. Four further things confuse people. input.sensitivity is clamped to the range -1.0 to 1.0, so cranking the number past 1 does nothing. input.* settings are global and hit the touchpad and trackpoint as well as the mouse. force_no_accel is read only as the global input:force_no_accel and cannot go inside a per-device block. And flat is not free: libinput describes it as "a constant factor applied to all device deltas, regardless of the speed of motion", so you lose the slow-precision and fast-travel split and usually need more DPI or more physical movement to cross the screen.
Fix
**1. Know what you are trading before you turn acceleration off.** libinput's default profile on every device is adaptive, which scales the pointer with how fast you move. flat is "a constant factor applied to all device deltas, regardless of the speed of motion". That is what you want for aiming, and the cost is that a slow careful movement and a fast flick now use the same factor, so crossing a wide desktop takes more physical travel. The usual answer is to raise the mouse's own DPI rather than the sensitivity. On a touchpad, where physical travel is small, adaptive is usually the better choice, which is one more reason to scope this per device.
**2. On Omarchy 4 the lines are already in the file, commented out.** ~/.config/hypr/input.lua ships as one fully commented-out hl.config block that already contains accel_profile and sensitivity. Uncomment only those two lines and the hl.config({ input = { ... } }) structure around them. Do not uncomment the whole block: it also carries kb_layout = "us,dk,eu" and a kb_options string, and kb_options replaces rather than merges, so uncommenting it silently drops Omarchy 4's compose:caps,shift:both_capslock_cancel.
-- ~/.config/hypr/input.lua
hl.config({
input = {
-- Turn off mouse acceleration (libinput default: adaptive).
accel_profile = "flat",
-- Clamped to -1.0 .. 1.0. Omarchy 4 already sets 0 in
-- /usr/share/omarchy/default/hypr/input.lua, so leaving this at 0 changes
-- nothing and is easy to mistake for a failed edit.
sensitivity = 0.35,
},
})
Omarchy 4 sets sensitivity = 0 and never sets accel_profile, so the acceleration you are fighting is libinput's own default and not something Omarchy chose.
**Plain Hyprland 0.55 or newer**: the same block goes in ~/.config/hypr/hyprland.lua.
**Hyprland 0.54 and older only** (hyprlang, ~/.config/hypr/hyprland.conf). Omarchy 4 never ships this:
input {
accel_profile = flat
sensitivity = 0.0
}
Apply with hyprctl reload.
**3. Try values live before committing them.** On 0.55 and newer:
hyprctl eval 'hl.config({ input = { accel_profile = "flat", sensitivity = -0.3 } })'
hyprctl keyword input:accel_profile flat is the 0.54-and-older form. On 0.56 the Lua parser refuses it with "keyword can't work with non-legacy parsers", so do not copy it from an older guide.
**4. Scope it per device so the touchpad is not collateral damage.** Hyprland lowercases the libinput name and turns spaces into hyphens, so do not guess it. Print the exact strings:
hyprctl -j devices | python3 -c 'import json,sys; d=json.load(sys.stdin); [print(m["name"]) for m in d["mice"]]'
Real names from a live Omarchy 4 machine: razer-razer-basilisk-v3, corsair-corsair-gaming-k55-rgb-keyboard-1.
-- flat and slower for the gaming mouse only
hl.device({
name = "razer-razer-basilisk-v3",
accel_profile = "flat",
sensitivity = -0.4,
})
-- leave the touchpad on the adaptive curve, tune scroll separately
hl.device({
name = "elan1200:00-04f3:30fe-touchpad", -- your own name from hyprctl -j devices
accel_profile = "adaptive",
natural_scroll = true,
scroll_factor = 0.4,
})
Touchpad options go in an hl.device block under their bare names, not under a nested touchpad = { ... } table. Hyprland resolves a per-device key by its bare name and picks the fallback by device type, so natural_scroll inside a touchpad's block falls back to input.touchpad.natural_scroll and inside a mouse's block to input.natural_scroll.
**5. If sensitivity is still not enough range**, do not fight the clamp. Change the mouse's own DPI, on the mouse itself or with piper and libratbag for supported mice, or define a custom curve. The form is custom <step> <points...>:
hl.config({ input = { accel_profile = "custom 200 0.0 0.5" } })
**6. Absolute last resort**, force_no_accel = true bypasses most pointer processing for the rawest possible signal. The wiki explicitly does not recommend it, because of cursor desynchronisation, and Hyprland reads it only from input:force_no_accel, so it cannot go inside an hl.device() block. It is all devices or nothing:
hl.config({ input = { force_no_accel = true } })
**7. Confirm the edit landed.** After hyprctl reload:
hyprctl configerrors
hyprctl getoption input:accel_profile
hyprctl getoption input:sensitivity
accel_profile reads str: [[EMPTY]] with set: false on a stock Omarchy 4 machine and must read str: flat with set: true once your edit is in force. getoption reports global values only, so a per-device override cannot be read back this way and has to be judged by feel.
Verify
hyprctl configerrors prints nothing after hyprctl reload. hyprctl getoption input:accel_profile returns str: flat with set: true, where a stock Omarchy 4 machine returns str: [[EMPTY]] with set: false. hyprctl -j devices lists each pointer with the exact name you matched on. Move the mouse slowly then quickly across the same physical distance: with flat the cursor travels the same screen distance both times. Confirm the touchpad still feels normal if you scoped the change with hl.device(). getoption reports global values only, so a per-device override cannot be read back that way.
Sources
- https://wiki.hypr.land/Configuring/Advanced-and-Cool/Devices/
- https://raw.githubusercontent.com/hyprwm/hyprland-wiki/main/content/Configuring/Advanced%20and%20Cool/Devices.md
- https://raw.githubusercontent.com/hyprwm/hyprland-wiki/main/content/Configuring/Basics/Variables.md
- https://raw.githubusercontent.com/hyprwm/hyprland-wiki/main/content/Configuring/Advanced%20and%20Cool/Using-hyprctl.md
- https://github.com/hyprwm/Hyprland/issues/7088
- https://wiki.hypr.land/Configuring/Basics/Variables/
- https://wiki.hypr.land/Configuring/Advanced-and-Cool/Using-hyprctl/
- https://github.com/hyprwm/hyprland-wiki/blob/main/content/configuring/core/config-options.md
- https://github.com/hyprwm/hyprland-wiki/blob/main/content/configuring/core/devices.md
- https://github.com/hyprwm/hyprland-wiki/blob/main/content/configuring/core/advanced-configuration/using-hyprctl.md
- https://github.com/hyprwm/Hyprland/blob/v0.56.2/src/managers/input/InputManager.cpp
- https://wayland.freedesktop.org/libinput/doc/latest/pointer-acceleration.html
- https://github.com/omacom/omarchy/blob/quattro/default/hypr/input.lua
- https://github.com/omacom/omarchy/blob/quattro/config/hypr/input.lua