The text to video workspace. The form fills the left column, the finished clip plays on the right.

The form and the result stay side by side. You can change one setting and generate again without losing the clip you already have.

Fill the left column top to bottom. The order on screen is the order to work in, because the model you pick decides which controls appear below it.

Choose Text to Video, the second entry. Changing the type rebuilds the form underneath rather than loading a new page, so moving between modes keeps your place.

Text to Video is the only entry in this list that needs no input file at all.

Five generation types. Text to Video is the second one.

WAN 3.0 is second, under Seedance 2.5. The red badge marks relaxed content filtering.

Those two figures are limits rather than suggestions. Selecting WAN 3.0 removes 4K from the resolution control below, because the model stops at 1080p.

Read the list before you commit. If the delivery needs 4K, Seedance 2.5 is the model to pick. If it needs a clip under 4 seconds at a 30 second capable model, WAN 3.0 is the only option here.

Describe the movement as well as the scene. A prompt that names a subject and stops leaves the camera to the model, and the model will invent one.

Name the shot size, the direction of travel, what holds still, and the sounds you expect. WAN 3.0 reads the audio cues from this same field.

The prompt field. The counter in the corner reads 0 / 2500.

Five frame shapes. 16:9 is the default and carries the highlight.

Aspect ratio is a composition decision taken before generation. The model stages a subject differently in a vertical frame than in a wide one, so a vertical cut and a wide cut of the same idea are two generations.

There is no ultra wide 21:9 tile here. If the delivery is 21:9, generate at 16:9 and crop, or use a model that offers the ratio directly.

Resolution and duration together drive the wait and the credit cost, and they multiply. A 30 second clip at 1080p is the most expensive combination WAN 3.0 offers.

Test at 480p. Confirm the motion reads correctly, then raise the resolution once the prompt is settled.

Three resolutions. 720p is selected when the form loads.

A continuous slider, so any whole length between 2 and 30 seconds is available.

This slider is the reason to reach for WAN 3.0 during iteration. Two seconds at 480p is the cheapest way to find out whether the model understood the shot.

Cost scales directly with the number you set. Leave it at 2 seconds while the prompt is still moving.

A seed is what makes an A and B test possible. Fix the number, change one word in the prompt, and the difference in the output comes from the word rather than from chance.

The dice button fills a new random number. Copy the seed of a result you liked before you change anything else, because the field resets when you reload the workspace.

Optional. Empty means random, a number means repeatable.

Reasoning Mode. The same control is called Thinking Mode in image to video.

Turn it on for a prompt that describes several things happening in order, or several people interacting. Those are the cases where the model has to decide what happens in the middle of the clip.

Leave it off for a single action or a single camera move. The planning pass adds time, and a short simple prompt gains little from it.

Name the sound inside the prompt. "Rain on a tin roof", "boots in wet mud", "a sword drawn behind the frame" are read alongside the visual description.

Turn it off when you plan to lay your own sound over the clip. Silence is a valid choice and it removes a variable from the generation.

One toggle. Audio is generated with the picture rather than added afterwards.

Credits required reads 650 for 720p at 5 seconds with audio on.

Read that number rather than estimating. It already accounts for the resolution and the duration you set, and it is the figure that will be deducted.

The clip appears in the player on the right when it finishes. Download saves the file, Delete removes it.

Connect a Prompt node to a Text to Video node, set its Model to WAN 3.0, then wire the output into a Result node. The generation node confirms the wire with a green "Prompt connected" line.

An amber line reads "Connect a negative prompt node (optional)". A negative prompt is a second Prompt node listing what to keep out of the clip, and that is where exclusions belong.

The node reads ~350 credits for this run. The Result node collects batch output.

The finished clip opens full width, with Download and Close beneath it.

Click the output thumbnail on the node to open the clip full width. Download saves the file and Close returns you to the graph.

The node keeps its result after it runs. Change one upstream setting and compare the new output against the previous one without rebuilding anything.

The form is shorter than the text to video one. There is no aspect ratio control, because the source image sets the frame shape.

Everything else carries over. Three resolutions, the same 2 to 30 second slider, a seed field, and two toggles at the bottom.

The image to video form. Credits required reads 650 for 720p at 5 seconds.

Image to Video is the first of the five generation types.

The form rebuilds immediately. An Image Source section appears at the top and a Last Frame Image section sits below it, in place of the aspect ratio tiles.

Prompt text survives the switch, so an idea written for text to video can be reused against a still without retyping.

Two other models in this mode reach 30 seconds. Seedance 2.5 does it at up to 4K, and Vidu Q3 Drama at up to 1080p from a script.

Vidu Q3 Drama also starts at 2 seconds, so the choice between it and WAN 3.0 comes down to what else you need. Vidu Q3 Drama adds a script field with named assets, and WAN 3.0 adds native audio, a seed field and the last frame input.

The image to video model list. Several entries here differ from the text to video roster.

Five ways in. Upload, a previous generation, an AI Actor, an AI Character, or a URL.

This image sets the frame shape of the finished clip. Crop it to the ratio you want before uploading, because there is no ratio control to correct it later.

A clean source helps. Clear subject separation, even lighting and few compression artefacts give the model more to track when it builds the motion.

This is the control for a continuation. Take the last frame of a finished clip, make it the first frame of the next one, and the two cut together cleanly.

Keep the two frames close in subject position, lighting and shape. A large gap between them produces an abrupt change in the middle of the clip.

Optional. Setting it fixes the composition the clip ends on.

The same prompt field, 2,500 characters, in both modes.

The model already has the appearance from your image. Describing the subject again spends characters that should go on the movement.

Write what changes over the clip. Where the camera travels, what the subject does, how the light shifts, and what stays where it is.

Match the resolution to the source. Animating a small image at 1080p adds cost without adding detail the source never had.

Run the first pass at 480p and 2 seconds. That combination costs 140 credits and tells you whether the still animates well at all.

The same three resolutions as text to video.

2 to 30 seconds, the same continuous slider as text to video.

A still that animates well for 5 seconds does not always hold for 30. Motion invented from a single frame drifts further the longer it runs.

Build up in steps. Confirm at 2 seconds, then 10, before committing credits to a full 30 second pass.

Replace the -1 with a whole number to make a run repeatable. The dice button beside the field writes a fresh random number for you.

A fixed seed is most useful here when you are testing the same still against several prompts. It removes one source of variation between the runs.

The field defaults to -1, which generates a new random seed on every run.

Thinking Mode. Text to video prints the same control as Reasoning Mode.

Two names, one control. Worth knowing if you move between the two forms and expect the label to follow you.

Turn it on when the prompt asks for a sequence of actions from the still, and leave it off for a single camera move.

Enable Audio behaves the same way in this mode. The soundtrack is produced with the picture and follows what the prompt describes.

Press Create. The Credits required line above the button carries the exact cost, and the clip replaces the placeholder in the player when it finishes.

The audio toggle, on by default in both modes.

Image Upload and Prompt both feed the Image to Video node. This run reads ~650 credits.

Wire an Image Upload node and a Prompt node into an Image to Video node, then send its output to a Result node. The generation node confirms both wires with green "Image connected" and "Prompt connected" lines.

Reusing one Image Upload node across several generation nodes is the point of building this in a graph. One source frame, several models or several settings, compared in one view.

The sidebar holds Aspect Ratio, Resolution, Video Duration, Seed, Enable Audio and Thinking Mode. Aspect Ratio is absent from the workspace form, so the graph is the place to reframe a still while you animate it.

The Seed row here is labelled "Seed (Optional - -1 for random)", which states the convention plainly.

The node sidebar. Aspect Ratio appears here and not in the workspace form.

The finished clip, with Download, Delete and Close below the player.

The output opens in the same full width viewer as the text to video graph. Download saves the file, Delete removes it from the graph.

The Result node keeps a batch, so four variants of one frame gather in one node rather than four.