How to Make a Music Video From Your Own Track for Free
Most "free AI music video" tools mean one of three things: a watermark, a 15-second preview, or a beat-synced slideshow of abstract shapes. This walkthrough is about something narrower — turning your own uploaded song into a vertical 9:16 video with scenes, characters and cuts, and finishing it without paying.
It is doable. It has limits, and they are specific rather than vague. Both are below.
What "free" covers, exactly
The free plan on lolq.ai is 120 credits and no signup for the first clip — you land in the studio as a guest and your project follows you if you decide to make an account later.
Credits are pinned to what generation actually costs, so the arithmetic is short:
| Step | Cost |
|---|---|
| Story and concept | 0 |
| Cutting the song into scenes | 0 |
| Storyboard sketch sheet | 2 |
| Character model sheet | 2 |
| Frames for one scene (first + last) | 2 |
| Animating one scene (Grok) | 2 |
| Assembling the clip | 0 |
A scene therefore costs 4 credits, and a three-minute song is about 30 scenes — 120 credits, one complete clip. Text steps are free because they run on a subscription we already pay for, so the entire creative half of the process costs nothing.
Two things to know before you start: the optional sketch sheet and the optional character model sheet are 2 credits each, and the balance in the header updates before every action, so nothing gets spent without you seeing the number first.
Step 1 — Upload the track and pick a look
Drop in an mp3. An instrumental works; lyrics are optional.
Then pick up to three styles out of fifteen. They blend into one look, and the first one you pick is the base — the others tint it. The catalogue runs from warm hand-drawn anime and Shinkai-style skies through photoreal film and noir comic to claymation, thread embroidery, and several found-footage looks built out of 90s camcorder grain and flash.
Two pieces of practical advice:
Pick the style for the song, not for the screenshot. A style that reads as beautiful in a still can read as busy across 30 cuts. Analog looks — grain, VHS, flash — are unusually forgiving, because the texture hides the small generative errors that a clean 3D render puts under a spotlight.
Use the comment field. It is where context goes that is not in the audio: the city, the era, who the character is, what should never appear. It costs nothing and it changes the whole storyboard.
Step 2 — Story and storyboard (free)
The model reads the length and the structure of the track and cuts it into scenes with timecodes, shot sizes and camera moves. You get a scene list, and next to it a sketch sheet of the whole storyboard so you can see at a glance whether the story holds.
This is the step to be picky at. Fixing the storyboard is free. Fixing a scene after it has been drawn and animated is not. Read every line and rewrite anything vague before you generate a single frame. In practice, most of the difference between a clip that works and one that does not is decided here, at zero cost.
Two shapes work at this length: one story arc across the song, or a run of unrelated punch frames held together by style. Both are legitimate. Mixed is usually not.
Step 3 — Characters (optional, 2 credits)
If a person recurs, set them up once: a description, your own photos, or a generated four-angle model sheet. The face and the clothes then hold across scenes, while props — the helmet, the glasses, the bike — live as separate attributes and get pulled in by name only where a shot needs them.
Skip this and every scene re-invents the person from the text description. Over 30 scenes that drift is very visible, and it is the number one reason AI music videos look like a mood board instead of a video.
Step 4 — Frames, two per scene (2 credits each)
Each scene gets a first frame and a last frame; the motion is generated between them. If one frame is wrong, you redraw that frame — not the scene — and you can attach your own reference image for composition or light.
The habit that saves the most credits: one change per scene. If the first frame is a wide shot of a street and the last frame is a close-up of a face in a different room, that is not one scene, it is two. Keep the delta small and the motion comes out clean; make it huge and the model melts one picture into the other. Longer explanation of the technique here.
Step 5 — Animate (2 credits a scene on free)
The free tier animates on Grok, which runs through a gateway we already pay for. That is what makes a free clip possible at all.
The honest limit: Grok animates the first frame only. It does not take a destination frame, so shots move away from your picture rather than travelling to a chosen end point. Cuts are looser, complex action does not survive, and motion is suggested rather than specified.
For a first clip, an idea test, or a dry run of a storyboard, that is genuinely enough. If you later want motion that starts and ends where you decided, that is what the paid engines do — Seedance 2 Mini is the cheapest of them at 22 credits a scene, and the comparison is here.
Step 6 — Assemble (free)
Approved scenes are glued to your own audio — the file you uploaded, untouched — into one vertical clip. There is a film-grain checkbox that lays 16 mm grain over the finished video in a single pass rather than frame by frame, so the picture does not flicker. Then the mp4 downloads.
Everything is 9:16 from step one, so the clip goes into Reels, Shorts or TikTok without reframing and without bars.
Getting the most out of 120 credits
Five things that consistently pay for themselves:
- Spend all your effort on the free steps. Story and scene breakdown cost nothing and decide most of the outcome.
- Shorter is stronger. A 60-second edit is 10 scenes, 40 credits. You get three attempts inside the free plan instead of one, and short vertical videos perform better anyway.
- One idea per scene. Also the cheapest way to avoid melted frames.
- Redraw the frame, not the scene. Two credits instead of four, every time.
- Watch the counter. Every action shows its price before you press it. Nothing is a surprise.
What we are not going to promise you
The clip is made by models, and they are not perfect. Hands go wrong. A face drifts. A camera moves the opposite way from what the storyboard asked. On the free engine, some of that is not fixable, only re-rollable.
What the free plan honestly gives you is a complete three-minute vertical video built out of your own track, made of scenes you chose, in a style you picked, with your own characters — assembled and downloadable, with no watermark and no card.
That is enough to find out whether the idea in your head survives contact with the screen. Which is the only question worth answering before anybody pays for anything.
Open the studio → · See the four steps · Plans and per-scene prices
Related: What an AI music video actually costs · Seedance vs Kling vs Grok · First and last frame explained