How to make a morph transition from two photos

A morph transition does not cut between two shots and it does not fade between them. It grows one into the other. For about a second you are looking at something that is genuinely both images at once, and that second is the entire reason the effect works — it is slightly wrong, and slightly wrong is what stops a thumb mid-scroll.

The good news is that you do not need to build it. You need to pick two photos well. Almost every bad morph is a bad pair of photos, not a bad render.

What the AI is actually doing

A crossfade lowers the opacity of one image while raising the other. That is why crossfades look cheap: for half a second you can see both pictures and neither reads clearly.

A morph is a different operation. The model looks for structure the two images share — the line of a jaw, the edge of a horizon, the mass of a body in the middle of the frame — and then generates the in-between frames that move along those shapes. Nothing is fading. Every frame in the middle is a real, fully rendered image that did not exist in either source photo.

That is why the pairing matters so much. The model is looking for shapes to travel along. Give it two images with nothing in common and it has to invent the path, and invented paths look like mush.

Picking the two photos

The rule that gets you most of the way there: similar composition, different content.

• Centre the subject in both. Morphs anchored off-centre drift, and the drift reads as an error rather than an effect.
• Match the framing. Same rough distance, same rough crop. A close-up morphing into a wide shot gives the model two different problems to solve at once.
• Keep backgrounds simple. A plain wall, a sky, a single-colour studio backdrop. Busy backgrounds morph into visual noise because there are too many competing shapes.
• Change one big thing, not five. Face to a different face. Person to an animal. Car to a building. One clear transformation reads; five simultaneous ones read as a glitch.

Pairs that reliably work well:

1. The same person, two completely different looks
2. A face and an animal with a similar head position
3. A product and the thing it is made from
4. One location in two seasons, shot from the same spot

Making it in AI VFX

The whole flow is three taps.

1. Open AI VFX and pick your start photo, then your end photo.
2. Choose Morph from the four effects.
3. Tap render. It takes about 30 seconds, and you get a push notification when the clip is ready — you do not have to sit in the app waiting.

From there you can trim it, stack it with other scenes, or add music in the built-in editor before exporting to your camera roll. Or you can export it exactly as it came back, which is what most people do.

Getting it right on the second try

If the first render is not what you wanted, the fix is almost always in the inputs:

• Mushy middle? Your two photos were too different. Move them closer in composition.
• Subject drifting off frame? One of the two was off-centre. Re-crop both so the subject sits in the same place.
• Background chaos? Crop tighter, or reshoot against something plainer.
• Change too subtle? The two photos were too similar. A morph needs something to transform into.

When to use something else

Morph is the loudest way to say "look at this". That makes it wrong for some clips.

If you want the change to feel invisible — a before-and-after, an outfit change, a time-of-day shift where the story is the point — use Smooth instead. If you want a hard beat you can cut music to, use Swing. If you want production value rather than a trick, use Skew.

Morph is for when the transformation is the content.