Tesla Light Show Audio
There is one failure that looks like a broken show and is really a broken audio file: it starts in time, then slides. Here is why that happens, the exact spec Tesla wants, and the two things that cause nearly all of it.
.fseq. If a show drifts, the file is almost certainly 48 kHz.1. The sample rate, which is the whole game
Tesla expects the show audio at 44.1 kHz. A 48 kHz file does not get rejected. It plays, sounds completely normal, and drifts steadily out of time with the lights.
That is the diagnostic. A show that is wrong from the very first beat is a different problem, usually mismatched files or the wrong export. A show that starts tight and is visibly late by the second chorus is a sample-rate problem, every time. The gap grows because the car is playing the audio at a rate the timeline did not expect, so the error accumulates rather than staying put.
The reason this catches people who did nothing wrong: 48 kHz is the default in a lot of software. macOS runs its audio system at 48 kHz, so anything decoded or bounced through the system rate lands at 48 kHz unless you go and change it. Video tools default there too, because 48 kHz is the film and broadcast standard. Music is the outlier at 44.1, which is exactly the value Tesla wants.
2. MP3 padding, the quieter cause
Even at the right sample rate, an MP3 can sit a little late. MP3 encoders add a short pad of silence at the very start of the file that was not in the recording, and different encoders add different amounts. The whole song therefore begins a few tens of milliseconds after you think it does.
A few tens of milliseconds sounds like nothing and is nothing, until a snare hit is supposed to land with a flash. Then it reads as sloppy. This is why Tesla recommends WAV and why we convert whatever you upload into a WAV before it ships: WAV has no encoder padding and plays sample-accurately.
Most write-ups that mention this blame "MP3 quality". That is not it. A 320 kbps MP3 has the same padding as a 128 kbps one.
3. The spec, in full
- Sample rate: 44.1 kHz.
- Format on the drive:
.wavor.mp3. WAV is recommended and is what we export. - Filename: identical to the
.fseqbeside it, extension aside.song.fseqpairs withsong.wav. - Location: both files inside a folder named
LightShowat the root of the drive. - Length: the format allows hours; shows people watch run one to five minutes.
4. What the builder accepts, and what it does with it
Upload MP3, WAV, M4A, AAC, OGG or FLAC, up to 50 MB. Whatever comes in is decoded, resampled to 44.1 kHz if it is not already there, and written out as a WAV next to the .fseq.
So the file that ends up on your USB drive is correct regardless of what you started from. The spec above matters most when you are assembling a drive by hand, or bringing a show from somewhere else.
5. Fixing a file you already have
If you have a 48 kHz file and want it fixed outside the builder, any audio editor will resample. In Audacity, set the Project Rate to 44100 at the bottom left and export as WAV. In ffmpeg it is one command:
ffmpeg -i song.mp3 -ar 44100 song.wav
Resampling is not lossy in any way you will hear. It is arithmetic, not re-encoding, and it fixes the drift outright rather than papering over it.
Common questions
Why does my Tesla light show drift out of sync?
Almost always the sample rate. Tesla expects 44.1 kHz audio, and a 48 kHz file plays perfectly happily while sliding steadily further from the lights. That is the signature: it starts in time and gets worse, rather than being wrong from the first beat. Re-export the song at 44.1 kHz and the drift disappears.
Why is my audio 48 kHz when I did not choose that?
Because a lot of software defaults to it. macOS in particular runs its audio system at 48 kHz, so anything decoded or exported through the system rate comes out at 48 kHz unless you say otherwise. Video tools default to 48 kHz too, since that is the standard for film and broadcast. Music is the odd one out at 44.1.
Should I use MP3 or WAV for a Tesla light show?
WAV, and not only because Tesla recommends it. MP3 encoders add a short pad of silence at the start of the file that is not in the original recording, and different encoders add different amounts. That padding shifts the whole song against the lights by a few tens of milliseconds, which is small but visible on sharp hits. WAV has no such padding and plays sample-accurately.
What audio formats can I upload to the builder?
MP3, WAV, M4A, AAC, OGG and FLAC, up to 50 MB. Whatever you upload is decoded, resampled to 44.1 kHz if it is not already, and exported as a WAV alongside the .fseq, so the file you put on the USB drive is always the correct format regardless of what you started with.
Where does the song itself come from?
You bring it. We distribute the light pattern and never the audio, so you supply your own legally owned copy of the song. That is why every show is bring-your-own-music, and it is also why the audio spec matters to you rather than being something we could quietly handle out of sight.