Repository navigation
Releases
v1.5.1 - One-Click Install
Star Wars Episode I Racer PCVR now installs in one click.
- Download Star Wars Episode I Racer PCVR Installer.exe below and press INSTALL. It finds your game (Steam, GOG or any folder), puts the mod in and checks your PC for anything that would stop VR from starting.
- Already on 1.5, or an earlier version you copied in by hand? The installer finds it and updates it. Your settings and saves are kept.
- New: a VR folder in the game folder with a launcher for VR, one for the monitor, Uninstall.bat and Collect logs.bat.
- The racing is exactly 1.5: same mod, same build.
Prefer the zip? Unzip it and double-click Setup.bat: it does the same, in a window.
The mods are free, always. If you would like to support them: https://www.patreon.com/GameOrDie
v1.5 - Smoother Cockpit
Star Wars Episode I Racer PCVR v1.5 - Smoother Cockpit
Cockpit view is much calmer on rough ground, and VR is now far easier to get running.
Five people couldn't start it, and not one of them had put the files in the wrong place.
Smoother cockpit view
On rough stretches, cockpit view could feel harsh and jarring in a way the rest of the race
doesn't. It was measured rather than guessed at: the pod's body rises and falls about the same
everywhere, but through those stretches its nose pitches up and down five to seven times a
second, and in cockpit view that went straight into your eyes.
Nose-bob smoothing steadies only that fast pitching. Steering is never touched, and hills,
ramps and jumps still follow the pod. It's on by default; set it to 0 in F5 for the v1.4 camera.
One-click setup
The archive now contains setup.bat. It finds the game through Steam's registry, in every Steam
library, or through GOG. It backs up whatever it replaces, copies the mod, and then checks your
PC before you ever put the headset on.
That check is the point. The four things that actually stopped people were a missing 32-bit
OpenXR runtime, Meta's runtime, a stale ReShade API layer, and the game running on the
integrated GPU. All four are visible from the desktop beforehand, in plain text, instead of
afterwards in a log file while wearing a headset.
You can re-run just the check any time, without copying anything:
powershell -ExecutionPolicy Bypass -File setup.ps1 -CheckOnly
When VR still won't start, the mod now tells you why
vr_report.txtappears next to the game's exe: a short file with only the VR startup
lines. The same lines were always inhook.log, around line 2,500, which is why two support
threads stalled trying to find them.- Error codes are decoded, and the advice matches the error.
XR_ERROR_GRAPHICS_DEVICE_INVALID
now explains GPU selection instead of talking about the headset service. The person who
hit it confirmed that was the fix. - Broken OpenXR API layers are detected and bypassed for this game only. A stale ReShade
layer fails every runtime identically, which is why swapping runtimes didn't help anyone
who had one. - A working 32-bit runtime is found automatically if the registered one can't serve a
32-bit game. It applies to this game only, and your system-wide setting is left alone.
Ground no longer disappears when you look down
The engine culls to a cone built from the pod's camera, and that cone is not round: it derives
its vertical angle from the horizontal one and the window's aspect ratio. On a 16:9 window,
with the game's own 45° field of view, the result was a 28° vertical cone, narrower than
the headset's own view. The engine was discarding track you could plainly see, along a dead
straight line, and a seated player looking down lost most of it.
The cone now follows your head. Engine cull FOV boost becomes a floor rather than the main
control.
Fixes
- Black results screen and hub after a race. After the finish, cockpit view could keep
driving the camera from a race that had already ended, reading your pod's position from memory
the game had freed. The camera ended up in the void and nothing rendered until the main menu.
Whether it happened depended on what was left in that memory, so it came and went. Cockpit view
now only takes the camera while the game's own camera is following your pod. - Cockpit view survives a death. It reverted to the chase camera after every respawn, and
no other view did. - Look back (X) works in cockpit view again. It broke when cockpit view arrived: cockpit
view replaces the game's camera, so the game entered look-back and the view never turned. - Menus: one stick direction at a time. A thumbstick near a corner counted as two
directions, so one flick moved a list and changed a setting. Adjustable, and switchable
off. Racing is untouched. - Settings save reliably, including all twenty-three per-pod seat positions.
New settings
dpi_unaware=0in the[vr]block opts out of the v1.4 display-scaling change.- Bump smoothing (cockpit view, off by default) filters the vertical jolt from a seam
in the track. It works by holding your head still while the pod moves, so at higher values
it can put your viewpoint inside the pod's own geometry. 0 is the v1.4 camera.
Two corrections to the v1.4 notes
Both were wrong, and both were in the README as well:
-
v1.4 said the display-scaling change left the headset image unaffected. It did not. The
eye texture is copied from the game's framebuffer, so on a scaled display that change
lowered headset resolution too. The trade was worth it on the machines where it was
measured, but it is a trade, and it should have been described as one.dpi_unaware=0
gives the old behaviour back. -
The README said the default world scale was derived from the pod's dimensions, because a
podracer is about 7 m. 7 m is one engine, not the whole vehicle, so the reference was
measuring the wrong part. The measurement is fixed and now reports itself inhook.log,
but the default is deliberately unchanged until there are enough real readings to set it
honestly. Treat World units per metre as a comfort dial, not a calibrated one.
Thanks
To everyone who sent a log instead of just saying it didn't work, and in particular to the
tester who filed nine numbered, reproducible issues with video for each. Four of them
are fixed above.
v1.4.1 - Race-End Crash Fix
Star Wars Episode I Racer PCVR — v1.4.1
Fixes a crash at the end of every race, introduced in v1.4. If you have v1.4, please update.
The crash
v1.4 dropped you to the desktop whenever a race finished — free play, tournament, all of it.
The cause was the per-pod cockpit seats, new in v1.4. To know which racer's seat position to
use, the mod reads the player's score record. At the end of a race the game frees that record
but leaves the pointer in place, so a check for "is this pointer empty?" passed and the read
went into freed memory.
It now checks that the race is genuinely still running before looking, and verifies the memory
is readable before touching it. A stale reading can now cost a slightly wrong seat position for
a single frame, and nothing worse.
Thanks to the player who reported it with a crash log attached — that log is what made it a
twenty-minute fix instead of a hunt.
Also in this release
- When VR fails to start, the log now explains why. If a 32-bit OpenXR runtime is
registered but refuses to start, the mod reads that runtime's own configuration, reports the
library it asks Windows to load, and says whether that file is actually installed — then names
the runtimes known to work. Previously this case said only that something had failed. - Runtime support, corrected. SteamVR 2.17+ is confirmed working by two users independently,
on a Rift CV1 and on a Quest, and is the route for any tethered headset. Meta's own PC runtime
does not work: it registers a 32-bit entry, so it looks correct, but fails to create a
session. If you are on Meta's runtime, switch to SteamVR — it takes one toggle in
Settings → OpenXR.
Nothing else changes. Same cockpit view, same twenty-three tuned seats, same performance, same
settings.
Install
Copy the contents of the archive into your Star Wars Episode I Racer folder, overwriting when
asked. Your settings and any seat positions you have tuned are kept.
v1.4 - Cockpit View
Star Wars Episode I Racer PCVR — v1.4
Cockpit view
The camera button now cycles five views instead of three:
Chase → Cockpit → Engine → Bumper → Far chase → back to Chase
Cockpit puts your head where the pilot's is, so the pod's cables and cowling sit around you
and the engines pull from ahead. It's on the button you already use — no menu, nothing to
configure.
Every one of the twenty-three racers has its own seat position, set by eye in a headset one
pod at a time. They genuinely differ: the tuned heights span nearly two game units and the
front-to-back positions span almost five, so no single offset could sit correctly in all of
them. If one isn't right for you, F5 → VR names the racer you're flying and gives you
Seat up / back / right — saved against that pod alone, with a reset to the built-in value.
Follow pod roll controls how much the camera banks with the pod: 1 welds you to it, 0 keeps
the horizon level. Lower it if the rolling is uncomfortable.
Performance
Windows display scaling no longer decides your VR render resolution. If your desktop runs
at 125%, 150% or anything other than 100% — normal on 4K monitors and most laptops — earlier
versions rendered at your monitor's full physical resolution and sent that to the headset. On
a 4K screen at 150% that's 8.3 megapixels per eye instead of 3.7, for no visible gain.
Measured on the development machine, same race, twenty minutes apart:
| v1.3 | v1.4 | |
|---|---|---|
| frame time left over | 0.25 ms | 2.89 ms |
| frames missing the 90 Hz budget | 7.3% | 1.0% |
Nothing to configure. If v1.3 ran worse for you than you expected, this is likely why.
Side effect: on a scaled display the desktop mirror window is now upscaled by Windows, so the
picture on your monitor is slightly softer. The image in the headset is unaffected.
Fixes
- ~28,000 OpenGL errors per session, eliminated. With anti-aliasing off — the default — the
renderer built its main framebuffer as a multisample surface holding a single sample, then
asked the driver to copy from it. Every one of those copies was refused, roughly 1.7 per
frame. Now zero. - 100 ms freezes from the debug logger. The OpenGL debug callback was writing to disk
synchronously on every message; a session could spend over 24,000 blocking writes there. - Settings that silently never saved. Several options worked all session and were gone on
relaunch, because the save routine's change-check had never been told about them. - When VR can't start, the log now says why, and names which runtimes provide the 32-bit
OpenXR support this mod needs. A tethered Rift user spent two days on this. - SteamVR is confirmed working from 2.17 onward, on a Rift CV1. That's the route for any
tethered headset — Rift, Index, Vive, WMR — which Virtual Desktop can't serve.
Install
Copy the contents of the archive into your Star Wars Episode I Racer folder, overwriting when
asked. Upgrading from v1.3 keeps your settings and your tuned seat positions.
v1.3 - Crash Fix & Performance
Two fixes that matter to everyone: a crash at the end of a race, and a stutter that was costing you roughly a tenth of your frames.
Nothing else changes. Same controls, same settings, same requirements: a 32-bit OpenXR runtime (Virtual Desktop works, SteamVR does not), and your own copy of the game. Everything in the v1.2 notes still applies — analog steering, haptics and wheel support are unaffected.
The stutter
If the game felt like it was hitching about once a second no matter what was happening — in menus as much as mid-race — that was real, it was our fault, and it's fixed.
The mod checks for a connected gamepad so the D-pad can drive the menus. That check was supposed to run once a second. Because of a mistake in the condition, the rate limit only applied once a controller had already been found — so if you didn't have a gamepad plugged in, which is most people in VR, it asked Windows about all four controller slots every single frame, around 360 times a second. Asking about an empty slot is expensive: it goes all the way down into Windows' device enumeration, and that enumeration re-runs about once a second. Hence a freeze, once a second, forever.
Measured on a Quest 3 before and after, same machine, same session length:
| before | after | |
|---|---|---|
| average | 83–87 fps | 90.0 fps |
| frames over budget | 7–36% | 0–3% |
| worst frame | 130–190 ms | 12–21 ms |
A 130 ms freeze at 90 Hz is a dozen frames gone at once, which is why it read as stuttering rather than as a low framerate — the median frame was always fine.
Two related improvements came out of the same work: the Gamepad navigation setting now switches off the checking as well as the feature (before, turning it off still cost you the full price every frame), and the renderer no longer re-reads both shader source files from disk every time it compiles a shader.
If you previously turned settings down to cope with this, try turning them back up. AI full LOD in particular was measured as costing nothing.
The crash
A player reported an instant crash to desktop every time a race finished, on every track, with a freshly verified install. Someone else on the same build couldn't make it happen once.
That difference is the diagnosis. When identical code crashes always for one person and never for another, it's usually reading memory that's already been freed — and what happens next depends on what the machine happened to put there afterwards, which differs from run to run.
v1.2 added a line of diagnostic logging that looked up the player's pod to print a setting once. It remembered where the pod was during the race; the game throws the pod away when the race ends; the input code kept looking there every frame while you sat on the results screen. It checked the pointer wasn't empty, and it wasn't — it just no longer pointed at anything real.
It was a leftover debug print that should never have shipped. It's gone, along with the bookkeeping that fed it.
Also in this release
The wheel centring-spring feature is removed. Across every log ever collected it had never once worked — every device, every session, reported "not a force-feedback device", the G923 included. The tension wheel users noticed came from G HUB's own Centering Spring in Non Force Feedback Games setting, which the README has always told you to enable. That still works and is still the way to get it. Its F5 checkbox is gone rather than left doing nothing.
Crash reports now identify themselves properly. The mod installs as dinput.dll, and Windows has its own dinput.dll, so a crash in either printed the same thing and the two were indistinguishable — which cost a day on the bug above. Reports now name the folder as well as the file, mark the mod's own frames, and record where it was loaded, so a report you send in can be traced to an exact line of code without anyone having to reproduce it.
Frame timing is now logged permanently — one line per 300 frames with the median, the worst frame and a dropped-frame count. If anything ever feels slow again, hook.log will say so in numbers.
Reporting problems
Grab hook.log from the game folder before relaunching — it's overwritten each launch. If it crashed, include the whole timestamped file from crashes/. Both fixes in this release came out of exactly that.
Licence
AGPL-3.0, inherited from SW_RACER_RE by tim-tim707 and contributors, on whose decompilation and OpenGL renderer this is built. Complete corresponding source is the pcvr branch of this repository. openxr_loader.dll is the Khronos Group's OpenXR loader, redistributed unmodified under Apache-2.0. The original game is the property of LucasArts / Lucasfilm.
v1.2 - Wheel Support
⚠️ Superseded — please update to v1.3.This version has two problems that are fixed there:
- A crash to desktop when a race finishes. It depends on your machine, so many people never saw it — but if it affected you, it happened every single race.
- Stuttering roughly once a second, in menus as much as in a race. Fixing it took the frame rate from 83–87 fps to a locked 90.
Also note: the centring spring described below never actually worked. The spring wheel users felt comes from G HUB's own setting, which the wheel section already tells you to enable.
Steering is properly analog now — the thing players reported most after v1.1 — plus controller haptics and experimental support for racing wheels.
No controls have moved since v1.1. If you already play, nothing to relearn.
Full setup, controls and known limitations are in PCVR_README.md. Requirements are unchanged: a 32-bit OpenXR runtime (Virtual Desktop works, SteamVR does not), and your own copy of the game.
Steering
Several players described the thumbsticks as "a dead zone, then full tilt, like a binary on/off", and noticed an Xbox pad felt fine by comparison. They were right, and it was not a tuning problem.
The mod steered by injecting arrow-key presses past a threshold. The game builds its steering value from those keys at full deflection, so there was no in-between to feel — while a gamepad drove the game's analog path directly. Two different code paths, which is why the two felt so different.
Steering now writes that value directly, with no threshold. Small movements give small turns.
Controller haptics
The Touch controllers now respond to what happens on track:
- bumping another racer, scaled by how hard
- scraping walls and terrain
- crashing
- a short kick when boost engages
Both strength curves were calibrated from logged data across several sessions rather than guessed, so a light scrape feels different from a heavy crash. Adjust or switch off under VR in the F5 panel.
About the game's Force Feedback screen: it reports NO FORCE FEEDBACK DEVICE DETECTED, and that is correct and unrelated. It lists DirectInput wheels and joysticks, and a headset is not one. Controller haptics are a separate path.
Racing wheels (experimental)
A wheel is detected automatically — the mod asks Windows what each attached device is, so a wheel enables wheel support and a gamepad does not. Nothing to switch on.
Steering, both pedals, the D-pad for menus, and the face buttons are all mapped, and everything is remappable under Input in the F5 panel. The mod also restores the wheel's centring spring, which the game switches off when it takes the device and never puts back.
Defaults are a Logitech G923 map. Other wheels report different axis and button numbers, so use the Input panel — turn or press a control and it lists whichever one moved.
Two settings on the wheel itself matter more than anything in the mod. In Logitech G HUB, enable Centering Spring in Non Force Feedback Games, and drop Operating Range from 900 degrees to about 240. A podracer wants a quick input, and a shorter range gives far finer control.
Tested on one wheel by one person. It should adapt to others, but expect to set the numbers yourself.
Also in this release
The F5 panel is reorganised. The VR settings are now called VR and appear first; wheel settings live under Input, where they belong.
Defaults now match a tuned configuration rather than first guesses — HUD size, overlay text size and menu placement. The download deliberately contains no ini, so the defaults are what a new player actually gets.
Known limitations
Unchanged from v1.1: overhead place numbers are suppressed, HUD and menus share one panel, frame rate is mid-to-high 80s on a 90 Hz headset (CPU-bound in a single-threaded 1999 engine rendered twice), and physics is framerate-coupled.
New, and specific to wheels: on a G923, Start, Back and the clutch pedal cannot be reached. They are not exposed to DirectInput at all — not by the game, and not by reading the device directly — so nothing here can bind them. Binding a keystroke to them in your wheel's own software works: Escape on Start gives you pause.
Reporting problems
Grab hook.log from the game folder before relaunching — it is overwritten each launch and opens with an environment block that answers most questions on its own. If it crashed, include the timestamped report from crashes/.
Licence
AGPL-3.0, inherited from SW_RACER_RE by tim-tim707 and contributors, on whose decompilation and OpenGL renderer this is built. Complete corresponding source is the pcvr branch of this repository. openxr_loader.dll is the Khronos Group's OpenXR loader, redistributed unmodified under Apache-2.0. The original game is the property of LucasArts / Lucasfilm.
v1.1 - Haptics & Fixes
Three fixes from player reports on v1.0, plus controller haptics.
If you already play v1.0, one control moved — see Controls that changed below.
Full setup, controls and known limitations are in PCVR_README.md. Requirements are unchanged: a 32-bit OpenXR runtime (Virtual Desktop works, SteamVR does not), and your own copy of the game.
Controls that changed
| v1.0 | v1.1 | |
|---|---|---|
| B | Slide | Back / pause — everywhere |
| Slide | B | Right thumbstick click |
| Look back | Right thumbstick click | X |
| X | nothing at all (see below) | Look back |
B is now a proper Back button. Two players reported having no way out of a race or a menu without reaching for the keyboard, or switching to flat mode to click Back with the mouse. B now backs out of every menu, and pauses during a race.
Slide moved to the right thumbstick click, which is the one thing to relearn. Look back moved to X.
X did nothing at all in v1.0. Two controller actions had been given the same display name, OpenXR rejected the second, and the button had been silently dead since the first release — it looked correctly bound in the source and simply was not. It is a real button now, and it is where Look back lives, so nothing else had to move to make room.
This is deliberately a fixed mapping rather than a context-sensitive one. The game exposes no reliable way to tell "a menu is open" from "you are racing", and two attempts to infer it both produced buttons that did the wrong thing in the wrong place. A button that always means the same thing is better than a clever one that is occasionally wrong.
Menu navigation
Either thumbstick now moves the selection, in both directions. Previously up and down only responded to the left stick, so reaching for the right one did nothing — reported as "trouble moving the selection to where I wanted".
Controller haptics
The Touch controllers now vibrate on impact, scaled by how hard:
- bumping another racer
- scraping walls and terrain
- crashing
- a short kick when boost engages
Both strength curves were calibrated from real sessions rather than guessed, so a light scrape feels different from a full crash. Tune or disable them under VR in the F5 panel: Controller haptics on impact, Haptic strength, and separate impact and wall scales.
About the game's Force Feedback screen: it reports NO FORCE FEEDBACK DEVICE DETECTED, and that is correct and unrelated. That screen enumerates DirectInput devices — steering wheels and joysticks with motors — and a headset is not one. It says the same on an unmodded install. Controller haptics are a separate path and do not depend on it.
Also in this release
A control that fails to bind now says so in hook.log. The X bug above was invisible because a failed binding was created silently; that cannot happen again.
Environment variables may not reach the game, so the switches now live in the ini too. The game usually launches elevated, and Windows gives an elevated process a fresh environment — so SWE1R_NO_VR=1 set in a shell or shortcut was being silently discarded. Both switches now also work as keys in the [vr] block of SW_RACER_RE.ini, which elevation cannot strip:
[vr]
no_vr=1 ; run flat, skip VR entirely
verbose=1 ; per-frame diagnostics
Prefer the ini keys.
Unchanged and still known
Overhead place numbers are still suppressed; HUD and menus still share one panel; frame rate is still mid-to-high 80s on a 90 Hz headset, limited by CPU draw-call submission in a single-threaded 1999 engine rendered twice; analog steering is still experimental and off by default; physics is still framerate-coupled.
Reporting problems
Grab hook.log from the game folder before relaunching — it is overwritten each launch. If it crashed, include the timestamped report from crashes/.
Licence
AGPL-3.0, inherited from SW_RACER_RE by tim-tim707 and contributors, on whose decompilation and OpenGL renderer this is built. Complete corresponding source is the pcvr branch of this repository. openxr_loader.dll is the Khronos Group's OpenXR loader, redistributed unmodified under Apache-2.0. The original game is the property of LucasArts / Lucasfilm.
v1.0 - Initial Release
Star Wars Episode I: Racer in VR — stereo rendering with true per-eye projection, head tracking, a world-locked panel for HUD and menus, and full Quest controller support. On a 1999 game with no available source, via a hook DLL.
Built on SW_RACER_RE by tim-tim707 and contributors. Full setup, controls, settings and known limitations are in PCVR_README.md.
Before you download
You need a 32-bit OpenXR runtime. This is the one hard requirement and it rules some setups out:
| Runtime | Works? |
|---|---|
| Virtual Desktop (VDXR) | Yes — ships and registers a 32-bit runtime |
| SteamVR | No — its OpenXR loader is 64-bit only |
| Meta / Oculus PC runtime | Probably — documented as supporting 32-bit apps, untested here |
The game is a 32-bit executable, so the runtime must be too. If yours has no 32-bit half, the game launches normally and runs flat, with the reason written to hook.log. To check: HKLM\SOFTWARE\WOW6432Node\Khronos\OpenXR\1\ActiveRuntime must exist and point at a 32-bit DLL.
You also need a legitimate copy of the game (GOG or Steam; developed against Steam). No game data is included in this download.
Installing
Unzip and drop dinput.dll, openxr_loader.dll and assets/ into the game folder next to SWEP1RCR.EXE. Start Virtual Desktop and connect your headset before launching.
Nothing else needs installing — dinput.dll is statically linked and imports only Windows system libraries. No Visual C++ redistributable, no MinGW runtime.
This also fixes the Steam release's launch crash. Verified on a clean machine: fresh Steam install crashed on launch vanilla; dropping these files in made it start and run. You do not need the dinput.dll that circulates on the Steam forums — this one takes the same slot.
What's in the box
Shaders and HD fonts only, about 440 KB. The community mod's assets/custom_tracks/ folder (~217 MB of user-made courses) is not included — it has nothing to do with VR. If you want it, take it from the SW_RACER_RE releases and drop it in beside the others; the two sets merge.
Known limitations
- Overhead place numbers are suppressed. The game stamps them as a screen-space effect during the HUD pass, which in VR is a flat panel, so they cannot anchor to the world. Drifting numbers are worse than no numbers.
- HUD and menus share one layer, so a size that suits racing can leave menus oversized. Separating them needs a second layer.
- Frame rate is mid-to-high 80s on a 90 Hz headset with strong hardware. The limit is CPU draw-call submission in a single-threaded 1999 engine rendered twice, not your GPU. 72 Hz is rock solid.
- Analog steering is experimental and off by default.
- Physics is framerate-coupled — the engine advances its simulation with the frame rate, which is why the community traditionally caps it at 24–30 fps. Higher rates work but change handling.
Lens flares are redrawn correctly as world-space geometry, fogged and depth-occluded per eye.
Reporting problems
Grab hook.log from the game folder before relaunching — it is overwritten each launch and opens with an environment block that answers most questions on its own. If it crashed, include the timestamped report from crashes/ too; those carry a symbolised stack. For anything obscure, set SWE1R_VR_VERBOSE=1.
Licence
AGPL-3.0, inherited from SW_RACER_RE. Complete corresponding source, including the VR layer, is the pcvr branch of this repository. openxr_loader.dll is the Khronos Group's OpenXR loader, redistributed unmodified under Apache-2.0. The original game is the property of LucasArts / Lucasfilm.