The Drop · September 20, 2026

Turn an MP3 into a video: cover art, one still image and no flashing

A song and a picture. It is the plainest thing anyone asks a visualizer for, and it is the one job the tool has to be talked down into doing.

Photo via Unsplash

Nobody searches for this because they want a video. They search for it because a bare audio file has nowhere to go. YouTube will not take an MP3, and a track with no picture attached cannot be posted anywhere that expects a rectangle. So the job is to wrap the song in the simplest possible frame and get on with your day.

The one line that already does this

Before any of the settings below, be straight about the alternative. If a completely static frame is all you want, ffmpeg does it without a browser, an account or a render that runs in real time:

ffmpeg -loop 1 -i cover.jpg -i track.mp3 -c:v libx264 -tune stillimage -c:a copy -pix_fmt yuv420p -shortest out.mp4

The two flags doing the work are documented plainly. The image2 demuxer describes its loop option as "If set to 1, loop over the input", which is what keeps the single picture coming for the length of the song, and -shortest is described by ffmpeg itself as "finish encoding within shortest input", which is what makes the video stop when the audio does.

Two things about that command catch people out, both worth knowing before you run it. The output takes the pixel size of your image rather than any standard video size, so a 1400 by 1400 piece of cover art gives you a 1400 by 1400 square video that every platform will pad out with bars. And libx264 refuses odd numbers: feed it a 1401 pixel wide image and it stops with "width not divisible by 2" and writes an empty file. Crop to an even size first.

The settings that make the app do it instead

This tool is built to flash, so turning an MP3 into a video with an image and nothing else means switching things off in a specific order. Seven changes:

1. Empty the text box completely. Text is never forced. The content picker only offers a text slot if there are lines in the box, so clearing it means no words ever appear.
2. Upload your one image in the Images section and leave its role on Any.
3. Image flash chance to 100, Shape flash chance to 0. Now your picture is the only thing the rotation is able to pick, every time it picks.
4. Flicker opacity to 0. That is a genuine off switch rather than a very low setting. At zero the flicker layer returns before it draws anything at all.
5. Trigger mode to Fixed interval, slider pushed to its maximum of 1500 ms.
6. Flash intensity down to 60 ms, its minimum.
7. Flash effect to anything except Strobe. Strobe is the harshest option in the list: it repaints the whole background in a palette color and alternates a black fill every 33 ms. Zoom punch and Shake are the mild ones.

What those last three add up to is worth the arithmetic. A 60 ms effect firing every 1500 ms means something is happening in about four percent of the runtime, at a rate of two thirds of an event per second. Flash rates become a seizure risk for people with photosensitive epilepsy somewhere above three per second, so this configuration is not near it. That stops being true the moment you shorten the interval or raise the flicker, which is the trade you are making when you do.

Square cover art in a rectangular frame

The render is 1920 by 1080, a ratio of 1.78. Cover art is square. Something has to give, and the app's default decides for you unless you tell it otherwise.

The default image fit is Fill frame, which scales your picture until it covers the whole canvas and crops whatever hangs over. On a square image that means it gets drawn 1920 wide and 1920 tall into a frame with room for 1080 of height, so you keep 56 percent of the artwork and lose the rest, split evenly off the top and the bottom. If the artist name sits near the top edge of the sleeve, it is gone.

Switch the fit to Fit whole image and the square lands at 1080 by 1080 in the middle with a 420 pixel black bar down each side. Neither of those is wrong, they are just different problems. Each thumbnail also carries its own fit dropdown, so a wide photo and a square sleeve in the same session do not have to share one answer.

If you care about the frame being full and the art being whole, the fix is upstream. Export a 16:9 version of the artwork with the square sitting on a background you chose, and upload that.

Still is not perfectly still

Two things happen on every flash even in this stripped back setup, and it is better to know than to wonder. The image gets a 12 percent push in at the peak of the flash and settles back over the flash duration, and a palette color washes across it at low opacity on an overlay blend. At 60 ms both are a blink. They are not nothing.

The other honest note is file size. Each render comes out as a 1080p WebM and the recorder targets 16 Mbps whether the picture is a strobing edit or a photograph that never moves, so a still image video from the browser is far bigger than the same thing from ffmpeg, where a frozen frame gives the encoder almost nothing to store. If you need a small file, convert after: the WebM and MP4 post covers what to convert to and when it matters.

So why open the app at all

If a frozen frame is the goal, use the command. The reason to open the visualizer instead is that a still image is a floor rather than a finish. The same upload, the same session, two sliders moved, and the picture is reacting to the track: put Trigger mode back on Beat detection and the image pushes in on the kick, add a second image and the rotation starts alternating them on the beat grid, bring the flicker up a little and you have something with a pulse that is still calm enough to sit behind an hour of music. None of that is available from a single encoded frame, and all of it is one session away from where you started.

The settings walkthrough goes control by control if you want to take it further than that.

The short version

• Empty text box, one image, image chance 100, shape chance 0.
• Flicker opacity 0, Fixed interval at 1500 ms, flash duration 60 ms, effect not Strobe.
• Default fit crops 44 percent off a square cover. Fit whole image gives you side bars instead.
• ffmpeg does the frozen version in one line, at the image's own pixel size, even numbers only.
• The app earns its place the moment you want the picture to move with the track.