A 30-second AI clip from your song lyrics: how it works

A 30-second AI clip is not a visualizer. It is a short story built around your song. The app reads the lyrics and the mood, offers three ideas, assembles a storyboard you can edit, and only then asks for tokens. The result is a 9:16 or 16:9 video built around your track: the cuts land on the beats, and the sound stays yours.

If you want a simple rhythmic picture without a story, Track video makes one for free in the browser. This guide is about the clip with a story.

What you get

• 30 seconds of video in 9:16 (1080×1920) or 16:9 (1920×1080).
• Your track as the sound. Your audio file is used as it is.
• Cuts on the beat. Scene changes land on the beats of the track, and each transition takes a quarter of a second.
• Two kinds of scenes. Some scenes are short video, the rest are pictures with slow movement.
• A file for 7 days. The finished MP4 stays on the order page for a week.
• An email when it is ready. It says the video was made with AI.

How to order: seven steps

1. Sign in and open the 30-second AI clip. You need an account from the first step: without one, the track cannot be analysed.
2. Upload your track as WAV or MP3, up to 50 MB. You can paste the lyrics: they are optional and limited to 3000 characters.
3. Type a mood or an idea if you like. Pick 9:16 for vertical video or 16:9 for YouTube.
4. Press "Analyse the track and suggest ideas". The app shows three ideas: what the song is about, who the hero is, how the story runs, the genre and the tempo. Pick one, or write your own.
5. Edit the storyboard. Each scene has a description, a type of "video" or "frame", and a length in seconds. Every scene must be filled in.
6. Look at the frames before you pay. The frames are drawn for free, but only if your balance covers the clip. One set of frames is available per draft, and another opens each time you change the idea.
7. Press "Create the AI clip". Tokens are charged only after you confirm.

What happens after you order

The order goes to the team for a check. The app says this usually takes up to a few hours. After approval the clip is rendered: some scenes are drawn by a neural network, and the rest is edited together with your track. If the render fails or the order is rejected, the tokens return to your balance and an email is sent.

How much it costs

An order costs 20 tokens. Tokens are sold in packs, and the bigger the pack, the cheaper each token. The price in your currency is on the pricing page, which always shows the current values, so this article does not repeat them.

What is free: reading the track, the three ideas, the first storyboard, editing the scenes, and the frame previews when your balance covers the clip. The first set of ideas costs nothing, and you can ask for three more sets per draft.

There is also a daily limit: orders are accepted within a limit shared across the site. If the limit is reached, the app says that orders are closed until tomorrow.

What the result looks like

An AI clip is not a film. The scenes are short, the movement is smooth, and the picture sometimes differs from the lyrics word for word. So look at the frames before you pay: if a scene does not work, edit the description and look again. Payment is taken once per order, after you confirm it.

The video is not a real recording. The email about the finished clip says it was made with AI. For what that means on release, read AI label before release.

What you cannot order

• Real people and celebrities. We do not make scenes with them.
• Adult content and violence. Such text fails the check before any tokens are charged.
• Scenes that need readable text on screen. A neural network does not guarantee legible lettering, so avoid it.

Showing the finished clip

The order form has a checkbox: you can allow the finished clip to be shown to other users as an example. It is off by default. Even with the box ticked, a clip appears in the examples only after the team marks it. Your name and email are never shown.

When to use the AI clip and when Track video

For a rhythmic video built from the sound and a title, use Track video: it is free and it is recorded in your browser during the fragment. For a story with a hero and scenes, use the AI clip. For a 15–30 second excerpt, you can use both: first the visualizer, then an AI clip on the main chorus.

How to read the ideas

Each idea is a short card: what the song is about, who the hero is, how the story runs, the genre and the tempo. The app built this reading from the text and the sound. If the lyrics were not recognised, the card says so and suggests pasting them for a more precise result.

Your own idea is a separate option. Describe the scene in your own words, and the app builds the storyboard around your description rather than around the suggested cards.

How to edit the storyboard

The storyboard is a list of scenes. Each scene has three fields: a description, a type ("video" or "frame") and a length in seconds. "Video" means a short moving clip, and "frame" means a picture with slow movement. Write descriptions concretely: the place, the action, the light. A vague scene gives a vague frame.

While at least one scene is empty, the order does not go through: the app asks you to fill in every scene.

What changes in the frames

When you change one scene and ask for the frames again, only the changed ones are redrawn. The rest come from the set that is already drawn, so editing one scene is cheaper than a full redraw. Each frame request uses one preview attempt for the draft, so do not change scenes over small details without a reason.

Frames are drawn in the format you chose for the order: 9:16 or 16:9. If you change the format, the frames are drawn again for the new one.

How much of the lyrics is used

You can paste the full lyrics, up to 3000 characters. The reading and the storyboard use the first 1500 characters. For a short song that is the whole text. For a long one, keep the most important lines near the top, because those are the lines the story is most likely to lean on.

If the frames do not render

Sometimes the preview fails. The app says that the frames did not render and gives the reason. Tokens are not charged for previews, because tokens are used only for the order. An attempt still counts toward the preview limit for the draft, so repeat it with care: fix the scene first, then request the frames.

What next

Try it. Open the 30-second AI clip, upload a track and get three ideas. Nothing is charged until you confirm the order. For a rhythmic video without a story, start with Track video.