You are browsing as a guest. Sign up (or log in) to start making projects!

Altaaf

@Altaaf

Joined June 20th, 2026

  • 8Devlogs
  • 5Projects
  • 1Ships
  • 20Votes
17yoam quite hungry
Open comments for this post
Reposted by @Altaaf

16h 19m 39s logged

BEAT IT

Your music is the level.

A synthwave motorcycle runner where the road, the collectibles, and the traffic are all generated from a song you actually own. Drop in any MP3, and the game listens to it — beat by beat — to build a highway that only exists because that specific track does.

What it actually does

Import any MP3. The game reads it, finds the beats, and figures out roughly where the melody rises and falls.
A road gets built to match: collectible beat blocks land exactly on the drum hits, and the path between them curves in the direction the melody moves.
Traffic weaves through the level too — some of it sitting directly on your line, forcing an actual choice between chasing a streak and swerving to survive.
Get clipped by a car and the screen reddens instead of ending the run outright. Get clipped enough times in a row without recovering, and it does.
Survive the whole song, and the camera holds still while the bike rides off toward the horizon on its own.

How to play

Controls: A/D or Left/Right to weave across the road.

Your music: On first launch, hit Play, then “Open Songs Folder” on the song select screen. Drop your MP3s in, back out to the main menu, hit Play again to refresh the list.

The story behind it

This started as a design doc for something I genuinely didn’t know how to build. I’d never touched audio analysis before — I knew Unity, I knew C#, and I had a pretty clear picture of the feeling I wanted (weaving through a highway that’s actually reacting to the song you picked), but the actual how of “read an MP3 and find the beat” was a total blank.

So I built it the only way that made sense: one small, testable piece at a time. Load the file. Confirm you can actually read the raw samples back out — and immediately learn the hard way that array indices and seconds-into-the-song are not the same scale when your audio is stereo and interleaved. Get a single number (energy in a window) that’s louder on a kick drum than on silence. Watch it also light up on vocals and melody, because raw loudness doesn’t know the difference between a kick and a chorus — and go learn what a low-pass filter actually does about that. Watch beats land audibly late, and find out that’s because you’re timestamping the start of a window instead of hunting for the actual peak inside it. Rebuild the whole detector around change in energy instead of raw energy, because a sustained bassline was getting picked up as a wall of false beats a flat threshold could never tell apart from a real hit.

The design itself changed shape more than once, too. The original plan kept the bike frozen in place with the whole world scrolling past it, specifically to dodge floating-point drift over a long run. Then a simple particle trail on the bike broke that illusion completely — a trail can’t read as “moving forward” if the thing it’s attached to never actually moves — so the bike started really driving, and the “drift” problem I’d been defending against turned out to not even apply once I did the math on how far a single song could actually carry it. A whole planned system (recycling road tiles forever) got deleted the moment I realized every level is scoped to one finite, fully-known-in-advance song — there’s no “forever” to build for.

The traffic system is maybe my favorite piece of this, honestly. Cars aren’t just scattered around — they’re placed by solving backwards: given exactly where and when I want a near-miss to happen, and how fast the car should be going, what starting position makes that car arrive there right on cue? Same trick the beat blocks use for their own placement, just pointed at a different problem.

Tech

No third-party DSP libraries — the beat and melody detection are hand-rolled: single-pole low-pass/band-pass filtering, spectral-flux-style onset detection, and a lightweight spectral-centroid approximation for melody direction, all built directly on raw AudioClip sample data

0
1
134
Ship Pending review

This is a rythm based Game that takes in MP3 Music files from your PC and turns it into a level through which you speed through on your bike.
It stated when i was playing “Dancing line” - a mobile game that also turns a song into levels although they are pre made and that makes it unadaptable to a song that i want to play the game with. In order to fix it I made a full audio analysis and beat detection system with local energy based beat detection.
It took a LOT of fine tuning to get the beats to be placed right for most songs but i feel it now does it reliably.
This is just V1 of the project next will be a fast fourier transform algorithm that will dissect an audio file to its absolute origins and get the more than enough data to build a good level.
Again this is the Demo for the project. Its still under development!
I hope you have fun playing it, read the Readme File or the itch.io description to get an idea of the controls!
Now go and BEAT IT

  • 1 devlog
  • 16h
Try project → See source code →
Open comments for this post

16h 19m 39s logged

BEAT IT

Your music is the level.

A synthwave motorcycle runner where the road, the collectibles, and the traffic are all generated from a song you actually own. Drop in any MP3, and the game listens to it — beat by beat — to build a highway that only exists because that specific track does.

What it actually does

Import any MP3. The game reads it, finds the beats, and figures out roughly where the melody rises and falls.
A road gets built to match: collectible beat blocks land exactly on the drum hits, and the path between them curves in the direction the melody moves.
Traffic weaves through the level too — some of it sitting directly on your line, forcing an actual choice between chasing a streak and swerving to survive.
Get clipped by a car and the screen reddens instead of ending the run outright. Get clipped enough times in a row without recovering, and it does.
Survive the whole song, and the camera holds still while the bike rides off toward the horizon on its own.

How to play

Controls: A/D or Left/Right to weave across the road.

Your music: On first launch, hit Play, then “Open Songs Folder” on the song select screen. Drop your MP3s in, back out to the main menu, hit Play again to refresh the list.

The story behind it

This started as a design doc for something I genuinely didn’t know how to build. I’d never touched audio analysis before — I knew Unity, I knew C#, and I had a pretty clear picture of the feeling I wanted (weaving through a highway that’s actually reacting to the song you picked), but the actual how of “read an MP3 and find the beat” was a total blank.

So I built it the only way that made sense: one small, testable piece at a time. Load the file. Confirm you can actually read the raw samples back out — and immediately learn the hard way that array indices and seconds-into-the-song are not the same scale when your audio is stereo and interleaved. Get a single number (energy in a window) that’s louder on a kick drum than on silence. Watch it also light up on vocals and melody, because raw loudness doesn’t know the difference between a kick and a chorus — and go learn what a low-pass filter actually does about that. Watch beats land audibly late, and find out that’s because you’re timestamping the start of a window instead of hunting for the actual peak inside it. Rebuild the whole detector around change in energy instead of raw energy, because a sustained bassline was getting picked up as a wall of false beats a flat threshold could never tell apart from a real hit.

The design itself changed shape more than once, too. The original plan kept the bike frozen in place with the whole world scrolling past it, specifically to dodge floating-point drift over a long run. Then a simple particle trail on the bike broke that illusion completely — a trail can’t read as “moving forward” if the thing it’s attached to never actually moves — so the bike started really driving, and the “drift” problem I’d been defending against turned out to not even apply once I did the math on how far a single song could actually carry it. A whole planned system (recycling road tiles forever) got deleted the moment I realized every level is scoped to one finite, fully-known-in-advance song — there’s no “forever” to build for.

The traffic system is maybe my favorite piece of this, honestly. Cars aren’t just scattered around — they’re placed by solving backwards: given exactly where and when I want a near-miss to happen, and how fast the car should be going, what starting position makes that car arrive there right on cue? Same trick the beat blocks use for their own placement, just pointed at a different problem.

Tech

No third-party DSP libraries — the beat and melody detection are hand-rolled: single-pole low-pass/band-pass filtering, spectral-flux-style onset detection, and a lightweight spectral-centroid approximation for melody direction, all built directly on raw AudioClip sample data

0
1
134
Open comments for this post

45m logged

02- Dimensions Corrected!!

So last time i made the model i though the number shown when i used the circle tool was radius BUT it was diametre so the dimesions were WAYY off.

Corrected them and used correctly sized models for the battery and arduino nano.

Im really tryna make it printable and i think its about done i’ll need to add the buttons and stuff tho lets see that tmrw tho.

Quote of the day:

“If you are waiting for the waiter…does that make you the waiter??”
~art of war (Sun Tzu)

0
0
3
Open comments for this post

51m 2s logged

LightSaber !!!

LightZaberrr

Why?:

  • So Other than all the technical things I really enjoy cosplaying and I did make Darth Vaders whole suit a while back with EVA foam,
    it was almost complete. Just Missing 1 thing

  • So yeah that is the motivation for it, coz buying a lightSaber is EXPENSIVE and imo making it is very much more cooler!

How?:

  • So what I’m thinking of rn is to first off have the most prominent feature i.e. The Blade Emerging, would be a light fill up
    Coz having a mechanically telescoping glade 😭 just STOP thats gonna be brutal.
  • then having an IMU to get the tilt and swing Data.
  • An Audio Module to process said data into the audio file’s playback parametres.
  • a Speaker!
  • I will have some controls for the Light colors but i think they should rather be dealt with remotely and not have more than one button on the lightsaber itself.
  • Made the Cad Model today and got the Dimension WRONG 😭😭, WHY WHY 😭so there goes my 1 hour .
0
0
7
Open comments for this post
Reposted by @Altaaf

1h 21m 51s logged

My Corner of the internet!-01

Why?:

Well first off its just a really cool thing to have, a personal website? that would look exactly like what i want ? hell yeah!!🔥🔥🔥

How?:

Probably only a static HTML/CSS/JS webpage as of now, I did learn about these languages in 8th grade (4 years ago💀) so mostly I brushed up on that and made basic stuff and used only placeholder texts other than in the about me (all that is true , yes ik im goated)

Why does it look like this?:

As you can see in the second image, Im using that color pallete for two reasons:

  • Its the album cover of one of my favourite band called “GlassBeach”.
  • The color pallete is lwk tuff on its own like its so unique but also not too much on the eye.

So yes thats it for today! Gonna make it Fire tmrw for sure for sure.

0
1
9
Open comments for this post

1h 21m 51s logged

My Corner of the internet!-01

Why?:

Well first off its just a really cool thing to have, a personal website? that would look exactly like what i want ? hell yeah!!🔥🔥🔥

How?:

Probably only a static HTML/CSS/JS webpage as of now, I did learn about these languages in 8th grade (4 years ago💀) so mostly I brushed up on that and made basic stuff and used only placeholder texts other than in the about me (all that is true , yes ik im goated)

Why does it look like this?:

As you can see in the second image, Im using that color pallete for two reasons:

  • Its the album cover of one of my favourite band called “GlassBeach”.
  • The color pallete is lwk tuff on its own like its so unique but also not too much on the eye.

So yes thats it for today! Gonna make it Fire tmrw for sure for sure.

0
1
9
Open comments for this post

51m logged

1 July 2026:

  • Submitted for re-review after making journal.md

2 July 2026:

  • GOT REGECTED 😭😭
  • Reviewer said the 3d model is lacking mounting spaces for electronics and i need a circuit diagram too.

3 July 2026:(2 hours)

  • Started working on making the 3d model better. (2 hour)
  • Got eepy went to sleep.

4 July 2026:(4 hours)

5 July 2026:

  • Welp Got regected lol. Reviewer said the model is vague about what goes where and there is no code.
  • Welp got the code structure done today.( 30 mins)

6 July 2026:

  • Got to work on the model , man onshape is toughh but its fun too like i can see it allows to get much advanced things done in it, took thier beginner course.

  • Added cavities and slot for the components while also making space for wiring to go through.

  • added spaces for servo’s horns to go in and get screwed into to get a firm hold of the legs.

  • yeah so going to send it for rereview, lets hope im not missing anything.

0
0
2
Open comments for this post

1h 5m 1s logged

Making It Real: TARS from interstellar

Why :

Ever since i watched interstellar, this particular robot has stuck with me. Its ability to make light of dire situations with its “sarcasm” setting 😭. There are a lot of other robots that i find cool, for instance Tony Stark’s robotic arm named “Dummy” Thats on my To do list too, But coming back to TARS I thought it would be better to make it as im only starting out, a robotic arm with AI would be quite a bold choice for a first hardware project.

How :

So originally TARS has 4 segment as you should be able to see in the images below, but having 4 segments is really complex, like say i want to rotate the rightmost segment, for that i would need a motor in the segment next to it , and when the motor rotates, it will not only rotate the rightmost segment, but would also rotate the segment that it is in, depending on which segment is heavier/ harder to rotate.

  • So im Simplifying TARS a LOT (Upgrades incoming in future for sure), like only having 3 segments (combining the middle two segments) and having all the components in it, so that when the motors try to rotate the left or right segment, they are rotated easily, also while TARS was autonomous my version would be running via remote control from a phone (with bluetooth or wifi connectivity).
  • I thought it would be cool to have a PID controller for its balancing so theres that too, I deliberately placed the accelerometer at the bottom of the middle section as velocity = angular velocity radius from point of rotation* .
  • thus the reading of accelrometer would be the Least amplified when its closest to the point of rotation i.e. the ground.

What will be used :

  • I would be using foam sheets for the body so that its all lighweight and easy to make too, I dont have a 3D printer ( that would be sooo convinient omg)
  • two metal gear Servos.
  • an Accelerometer
  • An Arduino nano
  • two 18650 li-ion batteries
  • 2.4 inch SPI TFT LCD display
  • Wiring Obviously
  • A soldering kit would be a blessing.
  • 5V buck convertor

The challenges :

What I feel is that this project is more mechanically challenged that it is in terms of coding and electronics.
TARS had a lot of ways to traverse, it would either start rolling by becoming sort of a wheel, or would have a pull itself forward by first lifting the outermost segments then placing them forward and then pulling them back.
To achieve that type of motion i would need one linear actuator per servo , to lift the outer segments up and then place them down when needed, that would make it unecessarily complicated.

  • what I did to dodge that is I Made the Legs slightly longer than the main middle segment.
    With that in rest position to be stable the segments should either be forward or behind the main segment.
  • I Pull both segments back slowly and the main segment moves forward ( due to high static friction between the legs and ground the main albiet heavier body is forced to move), then to have another stroke both the legs are SNAPPED foward VERY FAST, this causes the legs to go back in front while pushing the main segment back minimally as the kinetic friction is less and not strong enough to push back the main body.
  • Then I pull back to move forward again, and so the loop continues.
  • For Turning its the same stick-slip mechanism BUT only one leg at a time , thus TARS pivots about the resting leg!
    I found that pretty elegant. Although I DO plan to fully recreate TARS with Raspberry Pi and a custom AI agent and make it fully autonomous, It would have to wait until i first make this proof of concept.

Thank you for reading, Support me if you too find TARS cool and would like to help me start.

Peace OUT ✌

0
0
2

Followers

Loading…