Wan 3 is live on Skiprr
The new generation of Alibaba's video model generates from a prompt or from a starting image, at up to 1080p, for very few credits a second. Give it a frame you have already signed off and it moves that frame. It is in the model picker on every Skiprr video production today.
Alibaba released Wan 3.0, and Skiprr has added both tiers. This post is about Wan 3, the everyday one. Wan 3 Prime, the premium text to video tier, has its own announcement. Wan 3 works from a prompt alone or from a first frame, returns 480p, 720p or 1080p, and comes back with sound.
Start from the picture, not from a description of it
The most dependable way to get the shot you want from a video model is to stop describing the shot and show it the first frame. The product is already right. The label is already legible. The light is already the light you chose. All the model has to work out is what happens next.
We took a product still made earlier the same day and asked for very little: let the light shift, let the lavender stir, let a drop fall.
promptThe serum bottle stands on the stone as the window light shifts and the lavender stirs in a light draught. A single drop forms at the tip of the dropper and falls.
Watch the label as the camera closes in. It was set in the still, and it stays set: the same words in the same type, growing as the bottle does, without the shimmer that usually gives generated text away. The clip took the shape of the image it was given, which is why it is square.
Why it earns a place in the picker
It is inexpensive, and the price is simple. Each step up in resolution doubles the credits, so 480p costs a quarter of 1080p. Find the motion at 480p, then render the keeper once at the resolution the placement needs.
In Skiprr every scene of a video production is built on a composed keyframe, so a model that animates a frame faithfully fits the way the studio already works. The keyframe carries your product and your brand. Wan 3 carries it forward.
How to run it in Skiprr
Pick Wan 3 in the model picker on a video production. Work in image mode to animate a starting frame, or text mode to write the shot from nothing. Describe the scene, the light, the camera and the action in full sentences. Short prompts are expanded for you, and a negative prompt handles what you want kept out. The model will go longer, but scenes in Skiprr run up to 15 seconds each, which is where clips stay dependable. Seedance 2.5 stays preselected as the default, and your brand brain runs the show either way.