What the overlap claims
Every transition device makes a claim about the relationship between two shots, and the claim of a dissolve follows directly from its mechanics. For a moment the audience sees both images at the same time, superimposed, and superimposition means association.
That is why the conventional readings are time passing, memory, and comparison. A cut says next. A fade says over. A dissolve says these two belong together, and it is left to the context to say how.
The reading also depends on what is visible during the overlap. If the two frames share a compositional element, a matching horizon, a figure in the same part of frame, a shape that rhymes, the overlap looks intentional and the transition feels designed. If they share nothing, the middle of the dissolve is a mess of competing detail and the device reads as a default setting rather than a choice.
Duration and register
- Eight to twelve frames. A soft cut. Used to take the edge off a join that would otherwise be jarring, often between two shots of the same subject.
- Around one second. The classic passage of time. Reads as a small ellipsis.
- Two seconds and up. A statement. The overlap is long enough that the audience studies the double image, so the two shots had better be worth comparing.
- Very long, over three seconds. Belongs to montage and to dream sequences, where the superimposition is the content rather than the transition.
The reason the device acquired a dated reputation is that a whole generation of television and corporate editing applied a one-second version to every single join, whether or not time had passed. That is a problem of default settings, not of the technique, and the fix is simply to choose it when its meaning is the meaning you want.
Preparing shots for one
A dissolve is applied in the edit in a second, so nothing about it needs generating. What needs planning is the material on either side, and it has stricter requirements than a cut does.
The first is length. A one-second dissolve consumes roughly a second from the end of the outgoing shot and a second from the start of the incoming one, and during that time both are half-visible. Any action inside those windows is lost. Generated clips are often short and full of movement from the first frame, so the practical instruction is to generate the shots longer than needed, with dead air at the tail and head: held for six seconds, no movement except leaves.
The second is simplicity. Because both images coexist, complexity doubles. Static, sparsely composed frames overlap legibly; two moving crowds do not. Asking for a locked-off camera and a simple composition is worth more than any adjustment to the transition itself.
The third is compositional rhyme, and it is what separates a good dissolve from an adequate one. If you can generate the two shots with a shared element in the same part of frame, the overlap gains a visible logic. This is easiest to arrange when you generate both shots yourself, which is one of the few continuity advantages generated footage has over a location shoot.