How to make a lyric video in After Effects, start to finish
Most lyric video tutorials start with the typography. That is the fun part and it is not where the time goes. The order below is arranged around the one stage that eats a week if you let it, which is getting the words onto the right frames.
1. Set the comp up before you touch the type
Import the track first and build the composition around its duration rather than trimming to it later. Set the frame rate you are delivering at now, not at the end: changing it after you have keyframed text moves every keyframe onto fractional frames, and the drift shows up on exactly the fast lines where timing matters most.
If the video is going to more than one place, decide the aspect ratios up front. A 16:9 comp retimed to 9:16 is a re-layout, and a line that fitted on one line of a widescreen frame will wrap to three on a vertical one.
2. Get the words timed before you design anything
This is the stage worth reordering. The instinct is to design the look first and drop the words in afterwards, but the timing is what tells you how much room each line has. A line held for four seconds and a line held for four tenths are different design problems, and you cannot tell which is which until the timings exist.
There are three ways to get them. Tapping along to the track in real time is fast and imprecise, and you will redo it. Stepping through the waveform frame by frame is precise and slow, and it is the method most tutorials teach. Forced alignment does the pass automatically by matching your written lyrics against the vocal in the recording, which is what LyriSync does, and then you correct what it got wrong rather than placing everything by hand.
Whichever route you take, do it on the finished mix you are actually cutting to. Timings taken from a different master, a radio edit or a version with a longer intro will be out by a constant offset at best.
3. Decide the timing level before you build the layers
Per line, per phrase or per word is a structural choice, not a detail to settle later, because it decides how many layers you are about to make and what each one can do. A block of text that cuts in and out needs one cue per line. A travelling highlight, a per-word scale pop or type that assembles itself needs a cue per word.
Getting this wrong in either direction costs real time. Word level on a video that never animates individual words gives you hundreds of layers to manage for no visible benefit. Line level on a video that does animate them means going back and doing the whole timing pass again. Word-level vs line-level timing works through where the line sits.
4. Build one line, then repeat it
Design a single line completely: typeface, size, position, in-animation, out-animation. Get it exactly right, then turn it into the pattern everything else follows. Precomposing that one line and duplicating the precomp keeps a late change to the look from turning into two hundred manual edits.
If you are importing timings rather than placing them, this is the point where they arrive. Text layers that come in as real editable text can be restyled after the fact, so you can bring the timing in first and design on top of it. Anything that arrives as flattened images cannot, and it will force you back into the old order.
5. Animate against the vocal, not against the beat
A lyric video reads as tight when the text moves with the voice. Snapping everything to the grid looks correct in the timeline and feels mechanical on playback, because singers push and drag against the beat constantly and that is most of what makes a delivery sound human.
Give in-animations a short lead so the word is legible by the time it is sung rather than arriving after it. A few frames early reads as in time. A few frames late reads as broken, and it is the single most common reason a nearly-finished lyric video feels off without anyone being able to say why.
6. Check it at speed, and check it small
Watch the whole thing at full speed before you fix anything. Timing errors that are obvious when you scrub past them at a tenth of real time often disappear at tempo, and errors you cannot see while scrubbing jump out immediately. Fixing what does not need fixing is its own time sink.
Then watch it at the size people will actually see it. Type that is elegant at full screen is frequently unreadable on a phone, and the fix is a layout change, which you want to discover before the render rather than after it.
7. Render once, deliver many
Render a high quality master and derive the platform versions from it rather than re-rendering the comp for each one. If you kept the timing data as a file, you also keep the option of rebuilding the whole thing in a different look later without redoing the pass that took the longest.
Related: how to import SRT and timed lyrics into After Effects covers what actually lands in the comp, and syncing without an acapella covers getting timings from a finished master.