Wan 3.0 Pro: more creative freedom, video up to 4K

Wan 3.0 Pro gives you more freedom to create with relaxed content moderation, plus 2–30-second video generation with audio and output up to 4K. On API Stock, it starts at $0.086 per second at 480p. It is a useful new option when your creative brief is bolder than the safe, generic scenes you usually get from video generators.

Why is Wan 3.0 Pro interesting for creative freedom?

A video idea can lose its character before the first frame exists. You write a fashion scene, a moody music-video concept or an unsettling piece of fantasy, then spend your next attempts softening the description instead of improving the shot. For creators, that friction matters as much as resolution: the tool has to leave room for your actual brief.

The Wan 3.0 Pro route available through API Stock uses relaxed moderation. That is the central reason to try it for more adventurous creative work. You can describe the mood, wardrobe and atmosphere directly, then spend your attention on movement, lighting and composition. The public model is Wan 3.0 Pro, with the API identifier wan3.0-video-pro.

“Low moderation” describes the route's filtering approach. It is not a measured refusal rate, and we are not claiming that every possible prompt succeeds. We have not run a comparative moderation benchmark across video models. Requests can still be rejected, and the result can still miss the brief. Think of this as more room to explore your idea, rather than an acceptance guarantee.

That distinction lets you make a sensible choice: if creative latitude matters to your project, put Wan 3.0 Pro on your shortlist and test the actual scene you want to make. A generic landscape test will not tell you whether a model fits a distinctive fashion campaign or a dark fantasy sequence.

What kinds of scenes can you explore?

The strongest starting point is a clear visual idea, not a list of things the model should avoid. A summer fashion clip, an atmospheric horror teaser or a surreal music visual each needs its own camera language. Relaxed moderation gives you room to express the brief; a precise prompt gives that freedom a direction.

Seaside fashion. Imagine an adult woman in a light summer dress, wind moving the fabric, sea behind her and a slow camera move. You want an elegant campaign shot with a natural pose and warm light. Describe those details instead of filling the prompt with defensive wording.

An adult brunette woman, 28 years old, wearing an opaque flowing ivory summer dress beside the sea at sunset. A sea breeze moves her hair and the hem of the dress. She turns slightly toward the horizon. Medium shot, slow camera push-in, warm backlight, natural proportions, quiet waves and wind, no text.

The intended result is one fashion shot with a small, legible action. This is a prompt recipe, not a claim about a tested output. The cover illustrates the mood; it is a separately generated editorial image, not a frame rendered by Wan 3.0 Pro.

Dark fantasy. Build tension with an empty location, a strange sound and one reveal. An abandoned ballroom with moving shadows is more coherent than a request for a whole horror film in five seconds. Give the camera something specific to discover.

An empty coastal ballroom at night. Salt-stained curtains move in a breeze from an open doorway. One shadow crosses the far wall although the room is empty. Slow lateral tracking shot, pale moonlight, distant surf and a low creak of wood, no text.

Surreal music visuals. You can ask for an impossible environment without explaining away its strangeness: ocean waves suspended above a road, a changing sky, a dreamlike landscape. Start with a single transformation. Once it works, make a second shot and edit the sequence yourself.

These examples show creative directions to explore, not a catalogue of categories guaranteed to pass. Choose material you have permission to use, especially when your references contain a recognisable person. Greater creative latitude is most useful when you already know whose image and voice belong in your project.

How much does Wan 3.0 Pro cost?

The live API Stock catalog lists per-second pricing by resolution. Prices checked on 12 October 2026, in USD:

The examples multiply the listed rate by five seconds and round the total to cents. They assume no reference video. Reference-video seconds are billed in addition to output seconds, so a reference-based job can cost more than a text-only clip of the same output length. Check the current pricing before budgeting a batch.

Start short at 480p while you work on the scene. At the current rate, three five-second 480p drafts cost $1.29 in total; three five-second 4K attempts cost about $6.45. That difference buys room to refine the idea before you choose the final resolution. A higher-resolution rerun is a new generation, so it may also change the motion or composition; it is not simply an upscale of an approved draft.

What can you control besides the prompt?

Wan 3.0 Pro supports text-only generation, first and last image frames, and multimodal references. You can bring an image to establish the look, video to guide a scene, or audio as a reference. The image-to-video guide explains why starting with a still is useful when a particular outfit, character or setting matters.

The current public contract supports durations from 2 to 30 seconds, with a default of five seconds. Resolutions are 480p, 720p, 1080p, 2K and 4K; the default is 480p. You can choose a horizontal, vertical or square composition, among other ratios. Audio is enabled by default, and audio: false requests a clip without an audio track.

Frame inputs and reference inputs are different ways to direct a generation. Use a first frame when the opening composition should be anchored to your image. Use reference mode when the media should guide the generation more generally. Avoid combining a first/last-frame setup with reference-family inputs in one request.

For reference videos, the total source-video duration plus requested output duration must fit within 30 seconds. Keep this limit separate from the output-length limit when you plan a shot. The model documentation contains the full input contract and reference limits.

How do you start with one clear shot?

Open the model page, describe the scene and choose a short duration. Make the first attempt answer one question: does this framing and action work? Check the subject, wardrobe, camera movement and audio before increasing the resolution. If the model adds an unwanted action, simplify the brief instead of adding another paragraph of instructions.

For developers, the same workflow starts with a standard media request. Set API_STOCK_KEY to your own key in the environment:

curl -X POST 'https://api.api-stock.com/api/v1/generation/create' \
  -H "Authorization: Bearer $API_STOCK_KEY" \
  -H 'Content-Type: application/json' \
  -d '{
    "model": "wan3.0-video-pro",
    "input": {
      "prompt": "An adult brunette woman in an opaque flowing ivory summer dress beside the sea at sunset. A sea breeze moves her hair and dress. She turns toward the horizon. Medium shot, slow camera push-in, warm backlight, quiet waves, no text.",
      "duration": 5,
      "resolution": "480p",
      "size": "16:9",
      "audio": true,
      "promptExtend": false
    }
  }'

The request returns a task ID, not a finished video. Follow the documented task-status workflow to retrieve the result. promptExtend: false asks the service to keep your literal prompt instead of expanding it; it does not change moderation or guarantee exact adherence. You do not need to add a separate moderation switch to the public request.

When should you choose another model?

Choose Wan 3.0 Pro when relaxed moderation, reference inputs and flexible output length match your brief. If your priority is a different price point or a different visual style, compare the actual scene across models. Wan 3.0 is a separate catalog entry; its name alone does not promise the same filtering behaviour as this Pro route.

For strict brand work, creative freedom still needs human review. Inspect faces, hands, clothing continuity and the soundtrack before publishing. If you need a sequence with exact cuts, plan separate shots and assemble them in an editor. A 30-second generation gives you more time, but does not replace a shot list.

FAQ

Does Wan 3.0 Pro have low moderation?

The route offered through API Stock uses relaxed moderation and is intended to give creators more latitude. We have not published a measured refusal rate or a comparison proving it is the least restrictive model.

Can I generate content with no restrictions at all?

There is no guarantee that every request will be accepted or rendered as written. Relaxed filtering gives you more creative room, while service rules and rights to source material still apply.

Can I make video from a photo?

Yes. Use an image as a first frame, or select reference mode when it should guide the subject and style rather than define the exact opening frame.

Does it generate sound and 4K video?

Audio is enabled by default, and the catalog offers output up to 4K. You can also request a silent clip and choose a lower resolution for drafts.

How much is a five-second clip?

As of 12 October 2026, five seconds without video references cost $0.43 at 480p or about $2.15 at 4K. Reference-video inputs add billed seconds, so check the current pricing for your exact job.