Feature/speed streaks by sbilliet · Pull Request #5891 · supertuxkart/stk-code

sbilliet · 2026-10-08T16:46:21Z

Why

As discussed in the forums, I've been working on adding a speed streaks post processing effect to the game. The main purpose is to add to the player's experience of speed in the game.

Changes

The PR is split in two commits, in the first commit the notable things I've implemented:

The shader data/shaders/speed_streaks.frag, I've added extensive comments to explain how the shader works, but in a nutshell we convert UV coords into polar coordinates and then sample a single simplex noise (twice by using different UV scaling) and then layer them to extract noise peaks (a.k.a wisps). There are quite a few tweakable parameters, like density, softness, amount etc. that can be changed to fit the art direction. (I think I've done a pretty good job :) , but feel free to make changes if needed).
The post processing pass in src/graphics/post_processing.cpp passing in the main uniforms that control the shader u_intensity (0.0f - 1.0f value that controls the strength of the effect) and u_radial_scroll_offset (scrolling UV coordinates). As well as passing in a newly added seamless simplex noise texture (generated using FastNoise2 library) - to be added to the SVN repository. (See textures.zip)
Separate src/graphics/speed_streaks.hpp and src/graphics/speed_streaks.cpp class to neatly organize and store the dynamic intensity value and the UV offset for individual instances, as well as some static parameters that can be tweaked. This, like the other effects that are related to the player, is stored on the Kart. Ensuring it works for swapping cameras, split-screen, online etc. setups.
Added new accessors (getCurrentAdditionalSpeed and getCurrentAdditionalSpeedRatio) to src/karts/abstract_kart.cpp, I believe they have general utility in the codebase, but specifically to help calculating the boost intensity of the effect.
etc.

As for the second commit, as requested I've added a toggle to the video settings so that we can turn the effect on and off. I've followed suit of the other effects and added the relevant bits, like adding the CLI option and adding it to the in game benchmark (profiler). I've not yet added it to be automatically included in any of the graphics presets, to be discussed and added as a separate PR.

Agreement

By creating a pull request in stk-code, you hereby agree to dual-license your contribution as
GNU General Public License version 3 or any later version and
Mozilla Public License version 2 or any later version.

This includes your previous contribution(s) under the same name of contributor.

You also confirm that this contribution is your own original work
and that it does not contain code generated by AI tools.

Keep the above statement in the pull request comment for agreement.

Alayan-stk-2 · 2026-10-10T14:00:05Z

I haven't reviewed the code in detail yet, but I have been testing the feature.

The speedlines aren't particularly distracting for the most part. At max intensity, the speed of the streaks and the amount of screen they cover create quite a lot of movement so it is perhaps a bit much. When the kart is flying through the air after a zipper pad it doesn't matter, it'd be mostly for some situations with manual zipper use.
Just normal max speed + nitro is enough to activate the speed-lines. They are relatively faint at this speed, but in practice this means that good player who also drift will have streaks for most of the race.
I confirmed that getting the streaks to activate in Novice is fairly easy. I think even if we bump up slightly the activation threshold, it'll be fine. Note that the speed boost from nitro and zippers isn't scaled with difficulty in 1.x, so the % of generic max speed achieved with a nitro boost or a zipper boost is higher than in SuperTux. This compensate for drifting being rare.
The speed streaks are problematic on ultra-ultra-wide aspect ratios. This might seem unrealistic, but don't forget that playing in splitscreen with 2 players transform 16:9 into 32:9. The speed streaks take too much of the screen space and are relatively bigger.
The on-off graphics settings appears to work as intended.
The benchmark screen was already overcrowded (it was already overflowing with very big text size), now it's even more so. I think it's good to have the info displayed if possible, but this screen arguably needs a scrollable table on the model of what was implemented for soccer goals.
Performance cost on my machine is at about 45µs of GPU-time, so pretty cheap compared to most GPU settings (compare with MLAA at 300µs or bloom at 500µs with 800µs spikes).
I'm getting all 0 in the GPU profiler data from the benchmark, I had to do a manual profiler report to get the data. I would assume this issue is unrelated to this PR.

Regarding the commits, STK's repo is set to squash-and-merge (this avoids the noise from merge commits, and for most PRs it helps to make the history cleaner). I think this may be a case where it's actually cleaner for code history to have both separated.

So, once testing is complete and the code has been polished, it would make sense to have the video settings code in a separate PR.

SvenAndreasBelting · 2026-10-10T17:13:59Z

Just tested and I love the effect.

Imo it's in a great spot of helping with the feel of speed while not being distracting.

I haven't tried around with any parameter yet, I might do that at some point, but the effect looks great as is, so there are not really any modifications needed.

I agree with Alayan that the activation happens a bit too easy.

Alayan-stk-2

Here are all the things I noticed while checking the code.

The issue of the look of the effect in ultra-wide I mentioned previously also applies.

Also, speed streaks don't show up with ghost karts (probably because you can't properly probe their speed)

Alayan-stk-2 · 2026-10-16T14:20:43Z

+// ------------------------------------------------------------------------
+/** Returns the speed amount above the kart's own unboosted max speed.
+ */
+float AbstractKart::getCurrentAdditionalSpeed() const


Don't put any actual logic in AbstractKart.

AbstractKart is only inherited by Kart, so in the Evolution branch it has been fully removed as redundant and cumbersome.

In the 1.x codebase, you still need to get functions there so callers with an abstract kart can get things in and out, but actual logic here will make the future merging process into the Evolution branch more difficult.

Alayan-stk-2 · 2026-10-16T14:21:58Z

    {
        m_skidmarks->reset();
    }
+    if (m_speedstreaks){


Braces at the end of an if line don't follow STK's style conventions

Alayan-stk-2 · 2026-10-16T14:32:37Z

+        intensity            = speed_streaks->getBoostIntensity();
+
+        auto direction = SpeedStreaks::Direction::FORWARD;
+        if(cam->getMode() == Camera::CM_REVERSE)


The fact that you have the speedlines move in the right direction in reverse camera is a nice touch.

You can probably handle the assignment of direction with a simple ternary, it's cleaner.

Alayan-stk-2 · 2026-10-16T14:40:13Z

+{
+    World* world = World::getWorld();
+    Camera* cam = Camera::getActiveCamera();
+    const Kart* kart = cam ? dynamic_cast<const Kart*>(cam->getKart()) : nullptr;


This test is absolutely insufficient to know if the camera should actually have speed streaks displayed.

For example, the free camera (ctrl+F2) will still have speed streaks displayed. So does the top view camera. So do the various side of view cameras. Most of these cameras are logically tied to a kart, too.

Some kind of white list of camera modes where speed streaks make sense is probably the way to go.

Alayan-stk-2 · 2026-10-16T14:42:30Z

    }
+    if (m_speedstreaks){
+        m_speedstreaks->reset();
+    };


This ; does nothing and should be removed.

Alayan-stk-2 · 2026-10-16T14:45:18Z

+    const std::string noiseTexturePath = file_manager->getAsset(
+        FileManager::AssetType::TEXTURE,
+        "stk_noise_simplex.png"
+    );


The closing parenthesis on its own line isn't really something done in STK's codebase.

Alayan-stk-2 · 2026-10-16T14:54:27Z

+    float calcBoostIntensity(float lower_bound_ratio = 0.0f, float upper_bound_ratio = 1.0f) const;
+
+    // Reference to the kart to which these speed streak parameters belong.
+    const AbstractKart& m_kart;


In postprocessing, you actually already retrieve a Kart object rather than an AbstractKart object. So we should use Kart here too for the reason explained in my other comment (no AbstractKart class in the Evolution branch).

Alayan-stk-2 · 2026-10-16T15:03:15Z

+
+#include "graphics/speed_streaks.hpp"
+#include "karts/abstract_kart.hpp"
+#include <irrMath.h>


A blank line is typically left in STK's codebase between the local STK includes and includes from libraries.

Alayan-stk-2 · 2026-10-16T15:14:01Z

+
+    // Speed excess (as a fraction of the generic engine max speed) at which the
+    // streak intensity starts rising: 0.0 = the kart's own unboosted top speed.
+    constexpr static float s_boost_intensity_lower_ratio = 0.0f;


Before I saw this value in code, I thought the threshold was higher, because the effect at very low intensity is barely noticeable. This isn't a bad thing though, it means it fades in more naturally.

However, the speed streaks are a bit too easy to activate as mentioned in the previous review, so this should be increased. Based on my tests, I'd say 0.2f should be fine.

Alayan-stk-2 · 2026-10-16T15:16:33Z

+    constexpr static float s_boost_intensity_lower_ratio = 0.0f;
+    // Speed excess (as a fraction of the generic engine max speed) at which the
+    // streak intensity reaches 1.0f.
+    constexpr static float s_boost_intensity_upper_ratio = 0.75f;


I think this could be increased slightly, let's say 0.8f or 0.85f.

If you accumulate speed boosts, more than 0.75 is actually not that difficult to reach.

Alayan-stk-2 · 2026-10-16T15:21:29Z

How it looks at 5:1 ratio, and the speed lines get worse as they get closer to the outer borders.

21:9 splitscreen with horizontal split is about 4.7 ratio, so while extreme it isn't completely out of the realm of possibilities, and the degradation already happens at less extreme ratios like 3:1, it's just not as bad.

Alayan-stk-2 · 2026-10-16T15:39:03Z

How it looks at 5:1 ratio, and the speed lines get worse as they get closer to the outer borders.

A related issue is that the effect looks very weak at tall ratios close to 1:1.

sbilliet · 2026-10-16T22:18:40Z

Hey both thanks for the review. I had a bit of a busy one this week, but I should be able to address the feedback tomorrow.

qwertychouskie · 2026-10-17T04:50:19Z

The "Motion Blur" effect that's actually just a blur at the edge of the screen when activating a zipper should probably be either merged into this effect or removed entirely.

(An actual motion blur would be nice to have available in STK, but that's a different issue.)

sbilliet · 2026-10-17T11:12:21Z

@qwertychouskie I agree I think the motion blur could do with another pass. I was actually thinking after this PR merges I'm wanting to experiment adding proper motion blur to STK. We basically need to add an extra rendertarget to handle motion vectors and then change the post processing shader that's currently used for the radial motion blur and change it to use the motion vectors to handle the blur. Motion vectors are generally useful and can be used for a whole host of effects as well. Regardless we should start a separate thread on Motion Blur imo.

This reverts commit c3a63e6. Will be recommited in a separate PR for clarify.

sbilliet added 2 commits October 8, 2026 17:05

Implementation of speed streaks post processing effect

b8bd36d

Adding video settings toggle for speed_streaks effect

c3a63e6

Alayan-stk-2 mentioned this pull request Oct 16, 2026

refactor for speed observer #5609

Closed

Alayan-stk-2 reviewed Oct 16, 2026

View reviewed changes

Revert "Adding video settings toggle for speed_streaks effect"

d2fae5c

This reverts commit c3a63e6. Will be recommited in a separate PR for clarify.

Conversation

sbilliet commented Oct 8, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Why

Changes

Agreement

Uh oh!

Alayan-stk-2 commented Oct 10, 2026

Uh oh!

SvenAndreasBelting commented Oct 10, 2026

Uh oh!

Alayan-stk-2 left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Alayan-stk-2 commented Oct 16, 2026

Uh oh!

Alayan-stk-2 commented Oct 16, 2026

Uh oh!

sbilliet commented Oct 16, 2026

Uh oh!

qwertychouskie commented Oct 17, 2026

Uh oh!

sbilliet commented Oct 17, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

sbilliet commented Oct 8, 2026 •

edited

Loading