AetherWave Studio

THE ARENA

We review and compare models so you do not have to.

One prompt. One pair of reference images. One set of settings. Every model we host, generated the same way and reviewed side by side.

The premise

10s 9:16 vertical 480p 2 reference images no explicit first frame
Prompt, identical for every model Vertical wide shot of a massive cyberpunk megacity from a rooftop vantage, the protagonist's signal wave moving as a visible cyan aurora across the mid-level smog layer, weaving between broadcast antennas and billboard towers, growing stronger and more purposeful, the city grid lit below like a circuit board

Use case: a b-roll scene for a Track Reels music video. What matters here is not the prettiest single frame. It is whether the model can carry a look across many shots without repeating itself, because a full song needs twenty or more of these.

Carrier Wave cover art
Carrier Wave
The track these shots are cut to

Every scene has to survive being cut to this. A shot that looks fine alone and breaks the rhythm in sequence is a shot you cannot use.

The reference images

Both were supplied as references with the first-frame slot left empty. Every model ran 480p except Wan 2.6, whose cheapest tier is 720p, so it is shown there. Output dimensions vary a little between models because each interprets 480p its own way. That distinction is the whole test: some models treat a reference as a starting frame and hand it straight back, while others treat it as guidance and compose something new from it.

The results

Conditioning: composed a new shot from the references First-frame anchor: returned a reference as frame one Mixed: featured the subjects but re-lit the scene Not yet run

Run this exact test yourself

Everything above is reproducible. Take the same prompt, the same two reference images and the same settings, and see whether you get what we got. Ten seconds on the cheapest model on this page costs twenty cents.

Try it in Video Studio Or have us make the whole video

The test kit

Reference image 1 Reference image 2 Copy the prompt

Same brief, same images, same settings. If your results differ from ours we would genuinely like to know.