wan3.dev Paid content Multilingual

Wan 3.0 is a browser-based AI video generator that turns text, a photo, or an existing clip into a video of up to 1080p and 30 seconds. To make your first clip, pick the starting mode that matches the material you already have, write a prompt that names the subject, style, lighting, mood, and composition, then set resolution, duration, and audio before you generate. Expect a credit cost somewhere in the 98–5880 range, since the price scales with length and resolution.

Pick your starting mode

The generator offers four entry points. Choose based on what you have on hand, not on which sounds most advanced.

If you are starting from nothing, text to video is the fastest path to a finished result. If you already have a strong image, image to video usually gives you more control over the opening frame, because the model starts from your composition rather than inventing one.

Write a prompt that actually guides the model

The prompt field accepts up to 2000 characters, and the site's own tip is to be detailed and specific. A useful prompt covers five things:

• Subject — who or what is on screen
• Style — realistic, animated, cinematic, and so on
• Lighting — soft daylight, neon, backlit, overcast
• Mood — calm, tense, playful
• Composition — wide shot, close-up, centered subject

For example, instead of "a dog running," try "a golden retriever sprinting across a wet beach at sunrise, low camera angle, warm backlight, energetic mood, shallow depth of field."

If your idea is thin, use the Expand option next to the prompt field. It fills out a short idea into a fuller description, which is useful when you know the vibe but not the wording.

Set resolution, duration, and audio

Before generating, you choose:

• Resolution — 480p, 720p, or 1080p
• Duration — anywhere from 2 to 30 seconds
• Audio — whether to generate sound with the video
• Thinking mode — an option available in the generator panel

Higher resolution and longer clips both push the credit cost up. The full range runs from 98 to 5880 credits, so a short 480p test costs far less than a 30-second 1080p render. The generate button shows the credit price for your current settings (the example on the page reads 455 credits), so you can check the number before committing.

A practical approach: render a short, low-resolution version first to confirm the motion and framing, then re-run at higher settings once you like the direction.

Check your credits before you generate

Because cost scales with length and resolution, the same prompt can be cheap or expensive depending on settings. The generator displays the credit total on the generate button, so read it before you click. If you are experimenting with prompt wording, keep duration short and resolution low until the content is right — then spend the credits on a final high-quality pass.

The site lists a pricing page, so credit costs and any current offers are worth checking there rather than assuming a fixed rate.

Review and iterate

After the clip renders, judge it against what you asked for. If the motion is wrong, adjust the action wording in the prompt. If the framing is off, change the composition terms or switch to image to video and supply a starting frame that already has the composition you want. If the style is wrong, name the style more explicitly.

Switching input mode is often faster than rewriting a prompt repeatedly. A still image locks in the look; a reference video locks in the motion. Use whichever constraint your problem actually is.

What to know about the service

This is an independent, browser-based AI creation service. Its footer states that references to Wan 3.0 describe the creative product on the site and do not imply ownership, partnership, or endorsement by a model provider. All gallery videos and images are AI-generated synthetic content and do not depict real people or events unless explicitly stated.

Wan 3.0 on wan3.dev can animate a still image into a clip of up to 1080p and 30 seconds. Switch the generator to Image to video, upload your photo, then write a prompt that describes the motion and camera move rather than re-describing the picture. Expect a single run to cost hundreds of credits — the site's own example shows a 455-credit generate button — so decide your duration and resolution before you click.

Before you start

• Source image: a clear, well-lit still. The model animates what it sees, so a blurry or cluttered photo gives it little to work with.
• Prompt: up to 2000 characters. The site's own tip is to describe subject, style, lighting, mood, and composition — for image-to-video, add motion and camera direction on top of that.
• Credits: check your balance before generating. The interface shows "Available Credits" next to the generate button, and the example run is priced at 455 credits.
• Settings you'll choose: resolution (480p / 720p / 1080p), aspect ratio, duration (2–30s), and whether to generate audio.

Step-by-step: image to video

1. Open the generator and pick Image to video

The site lists four starting points: Text to video, Image to video, Video to video, and Video editing. Choose Image to video so the generator treats your upload as the first frame rather than as a loose reference.

2. Upload your image

Select the photo you want to animate. This becomes the visual anchor — the subject, framing, and lighting of the output should stay recognisable to it.

3. Write a motion prompt

This is where most first attempts go wrong. Don't describe what's already in the photo; describe what should change.

Weak prompt: "A woman standing in a field."

Useful prompt: "The woman turns her head slowly toward the camera, hair moving in a light breeze; slow push-in, warm late-afternoon light, calm mood."

Cover four things:

• Subject motion — what moves, and how fast
• Camera move — push in, pull out, pan, orbit, or hold still
• Lighting and mood — keeps the clip consistent with the source image
• Style — only if you want the output to shift away from the photo's look

4. Set resolution, ratio, and duration

Duration runs from 2 to 30 seconds and resolution from 480p to 1080p. Longer and higher-resolution clips are the expensive end of that range, so a first test at a short duration and lower resolution is the cheaper way to check whether your prompt produces the motion you want.

5. Decide on audio

The generator has a Generate audio toggle. If you don't need sound, leaving it off keeps the run simpler; if you do, the site advertises audio generation from text descriptions.

6. Check the credit cost, then generate

The generate button displays its price before you commit — the on-page example is 455 credits. Credit cost scales with the settings you chose, so if the number is higher than you expected, shorten the clip or drop the resolution and check again.

7. Review and adjust

Watch the result against your source image. Two common failure modes:

What it costs and what you get

• Output: up to 1080p, up to 30 seconds, from text, image, or video input.
• Model options: the generator lists Wan 3.0 with a 2–30s range at 480p/720p/1080p and a credit band of 98–5880 credits depending on settings.
• Pricing page: wan3.dev/pricing is linked from the main navigation and currently shows a 50% OFF label. Check it for current credit packs rather than assuming a price.

Practical notes

• Test cheap, finish expensive. Run a short, low-resolution version to validate the prompt, then re-run at your target settings once the motion looks right.
• One action per clip. A single clear movement reads better than a list of instructions, and it's easier to debug when the output is wrong.
• The site is independent. Its footer states that references to Wan 3.0 describe the creative product on the website and don't imply ownership, partnership, or endorsement by a model provider. Treat it as a browser-based creation service, not the model's official home.
• Gallery as reference. The workflow showcase on the homepage demonstrates subject action, camera direction, image-led motion, and visual transformation — useful for calibrating how much motion a prompt should ask for.

Wan 3.0's prompt box accepts up to 2,000 characters and its own tip tells you to describe the subject, style, lighting, mood, and composition. That is the whole method: name what is on screen, say what it does, say how the camera behaves, then set the light and the look. Prompts written this way work for all three starting points — text, a photo, or a reference clip — and you can adjust one element at a time when a result misses instead of rewriting from scratch.

The prompt pattern

Build every prompt in this order. Later elements matter less if the earlier ones are vague, so fix the top of the list first.

1. Subject — who or what is on screen, with enough detail to be specific ("a woman in a yellow raincoat," not "a person").
2. Action — the single main thing that happens during the clip.
3. Camera — how the shot moves or holds (static, slow push in, tracking alongside).
4. Lighting — time of day and light quality (overcast morning, warm lamp light, hard midday sun).
5. Mood and style — the feeling and the visual treatment (documentary, soft and quiet, high contrast).
6. Composition — framing and where the subject sits in frame (wide shot, centered, subject on the left third).

A working example:

A woman in a yellow raincoat walks toward the camera along a wet pier, slow push in, overcast morning light, calm and slightly melancholic, wide shot with the subject centered, muted color palette.

Compare that to "a woman walking on a pier, cinematic, 4K, beautiful." The second prompt gives the model nothing to act on — no direction of movement, no camera instruction, no light. Vague quality adjectives are the most common reason a clip comes back looking like a generic stock shot.

Adapting the pattern to each input type

The same six elements apply, but what you should spend characters on changes with how you start.

For image-to-video, don't re-describe the photo. If the image already shows a person at a desk, write the motion: "she turns her head toward the window, camera slowly drifts right, late afternoon light." Describing the subject again wastes characters and can push the model away from the source image.

For video-to-video, be explicit about the transformation. "Keep the camera movement and timing; change the setting to a snowy street at night, cool blue tones" tells the model what is fixed and what is not.

How much to write

The box holds 2,000 characters, but length is not the goal — coverage is. A prompt that names all six elements usually lands between 40 and 90 words. Below that, you are probably missing camera or lighting. Far above it, you risk contradicting yourself, which is worse than saying less.

Two habits that keep prompts tight:

• One main action per clip. If you want a character to walk in, sit down, and open a laptop, that is three beats competing for a clip that may run as short as a few seconds. Pick the beat you actually need.
• Cut adjectives that don't change pixels. "Stunning," "amazing," and "masterpiece" don't tell the model what to render. "Backlit," "shallow depth of field," and "handheld" do.

Matching prompt detail to your settings

Your prompt and your settings have to agree, or the result will feel wrong even when the model did what you asked.

• Duration. A prompt describing a slow, multi-stage action needs room to play out. On a short clip, describe a single moment instead of a sequence.
• Resolution. Fine detail you ask for — text on a sign, small objects, intricate patterns — is more likely to survive at higher resolution. At lower settings, favor larger shapes and simpler compositions.
• Audio. If you enable audio, say what should be heard. Ambient sound, dialogue, or music are different requests, and silence is a valid one to state.
• Aspect ratio. Write the composition to match the frame. A "wide establishing shot" fights a vertical frame; "centered subject with headroom" suits it.

When the result is wrong, change one thing

Rewriting the whole prompt after a bad clip makes it impossible to know what fixed it. Instead, diagnose by category:

• Wrong subject or wrong action → the top of your prompt is too vague. Add one concrete detail.
• Right subject, wrong feeling → lighting and mood are underspecified. Name the light source and time of day.
• Static or awkward motion → you described a scene but not a camera. Add an explicit camera instruction.
• Composition is off → state the framing and where the subject sits in frame.
• It ignored part of the prompt → you likely have two competing actions. Cut to one.

Change one element, regenerate, and compare. Two or three passes usually get you to a prompt worth saving and reusing with a different subject.

A reusable template

Fill this in and you have a complete prompt:

[Subject with one distinguishing detail] [does one action], [camera instruction], [lighting and time of day], [mood], [framing and subject placement], [style or color treatment].

Keep your best prompts in a note. Because the pattern is modular, you can swap the subject and keep the camera, lighting, and style — which is the fastest way to get a consistent look across several clips.

Wan 3.0 charges credits per generation, and the price of a clip is quoted on the Generate button before you confirm. On the site's generator page, the model selection line lists a range of 98–5880 credits for Wan 3.0, alongside settings of 2–30 seconds and 480p/720p/1080p. Your live "Available Credits" balance sits next to that quoted cost, so you can compare the two before spending anything. The practical rule: the longer the clip and the higher the resolution, the closer you land to the top of that range.

Where the credit cost comes from

Credits are the in-app currency for this independent browser-based service. You don't buy a fixed number of videos — you spend credits per generation, and the amount depends on the settings you choose at that moment.

The generator page shows three things side by side:

The 455-credit figure in the page evidence is the quoted cost for one specific configuration — not a fixed price for every clip. Change the duration or resolution and that number moves.

What pushes a clip toward 5880 credits

Two settings dominate the cost:

• Duration (2–30 seconds). A 30-second clip is roughly fifteen times the length of a 2-second one, and the top of the credit range reflects that.
• Resolution (480p/720p/1080p). Higher resolution means more frames to render at greater detail, so 1080p sits well above 480p.

Optional features can also change the total:

• Generate audio — the text-to-video panel includes an audio toggle, and the site describes generating "videos with audio."
• Thinking mode — listed as an available option on the generator panel.

Because these are optional, the safest habit is to read the credit figure on the Generate button after you've set duration, resolution, audio, and thinking mode — not before.

A workflow that keeps credit spend predictable

1. Write your prompt first. The prompt field allows up to 2000 characters, and the site advises describing subject, style, lighting, mood, and composition. A vague prompt that produces the wrong clip costs you credits twice.
2. Set duration and resolution before anything else. These two drive most of the cost.
3. Decide on audio and thinking mode. Toggle them on or off and watch the quoted figure update.
4. Read the Generate button. That number is your actual cost for this render.
5. Compare against Available Credits. If the balance is short, the Pricing page is the top-up route.
6. Test cheap, then commit. Draft an idea at a short duration and lower resolution, confirm the motion and framing work, then re-run at 1080p or a longer length.

Step 6 is the single biggest lever on total spend. A 1080p, 30-second render with audio is the most expensive configuration the page describes; the same idea validated first at low settings costs a fraction of that.

Common questions about the credit range

Why does the site list a range instead of one price? Because cost is a function of your settings, not a flat fee. 98 credits represents the cheapest combination the page supports; 5880 represents the most expensive.

Is the 455-credit figure the standard price? No. It's the quote shown for one configuration on the page at the time it was captured. Your own quote depends on your settings.

Do failed or abandoned generations cost credits? The available page evidence doesn't state a refund or failure policy, so treat the quoted cost as what you commit when you press Generate.

Does the balance carry over or expire? Not addressed in the available material.

Is there a free tier? The page shows an Available Credits balance and a Pricing page with a 50% OFF signal, but the evidence doesn't state whether any credits are granted free or whether login is required to generate. Don't assume either.

What to check before you generate

• The Generate button figure — this is the only number that matters for your next clip.
• Duration and resolution — the two settings that move cost the most.
• Audio and thinking mode toggles — optional add-ons that can raise the total.
• Available Credits — if it's below the quoted cost, top up via Pricing first.

If you're new to the tool, start with the shortest duration and lowest resolution that still tests your idea. Once the motion, camera direction, and subject look right, re-run at the quality you actually need. That sequence keeps you near the bottom of the 98–5880 range while you iterate, and reserves the expensive end for clips you already know work.

Reference to video in Wan 3.0 means starting from a clip instead of text or a still image: you upload a reference video, then describe what the new clip should inherit (subject action, camera movement, style) and what it should change (scene, wardrobe, lighting). On wan3.dev this sits under the Video to video start option, alongside Text to video, Image to video, and Video editing. It fits you best when you already have footage whose motion or look you want to reuse but whose content you want to replace. If you only have a written idea or a single photo, text-to-video or image-to-video is the cheaper, simpler path.

What the reference actually controls

Wan 3.0's generator lists three ways to start — Text / Image / Video — and the site describes the workflow showcase as covering "subject action, camera direction, image-led motion, and visual transformation." In practice, a reference clip is your strongest lever on:

• Motion — how the subject moves and how fast the action reads
• Camera language — pans, pushes, handheld feel, framing changes
• Style and grade — color, contrast, texture, overall look
• Transformation direction — what the clip turns into over its duration

What it does not reliably lock down is identity and exact detail. A clean reference gives the model a clear motion signal; a busy or low-quality one gives it noise, and the output drifts.

Preparing the reference clip

The site's stated output ceiling is up to 1080p and up to 30 seconds, so treat that as the target range for what you feed in.

Practical prep: trim to one continuous shot, keep the subject large enough in frame to read, and avoid clips where the camera and the subject both move hard at once. If your reference is longer than the clip you want, cut it to roughly the length you plan to generate.

Writing the prompt: separate "keep" from "change"

The prompt field on the generator accepts up to 2000 characters, with a tip to describe subject, style, lighting, mood, and composition. For reference-to-video, split your prompt into two explicit halves:

1. Inherit — name the motion and camera you want carried over: "keep the slow push-in and the walking pace of the reference."
2. Replace — name everything else: new setting, wardrobe, time of day, color palette.

Example prompt structure:

Keep the reference's forward tracking shot and the subject's steady walking rhythm. Change the setting to a rain-soaked night market, swap the outfit to a red jacket, shift the grade to cool blue with warm practical lights, keep the framing centered.

Vague prompts like "make it like the reference but different" give the model nothing to hold onto; the output tends to copy the reference too literally or ignore it entirely.

Settings and cost before you generate

The generator exposes model selection (Wan 3.0), resolution ratio, duration, a generate-audio toggle, and a thinking mode. The listed range is 2–30 seconds at 480p / 720p / 1080p, costing 98–5880 credits depending on those choices — so both length and resolution move the price, and the low end of that range is a short low-resolution clip while the high end is a long 1080p one. The interface shows a live credit figure next to the Generate button (the page displays an example of 455 credits for one configuration), so set duration and resolution first and read the number before committing.

For a first reference-to-video attempt, generate short and at a lower resolution to confirm the motion transfers the way you expect, then re-run at 1080p once the prompt is right. That keeps failed attempts cheap. Note that pricing and any discounts are shown on the site's own Pricing page — check there rather than assuming a rate.

When the output drifts from your reference

Most failures fall into three buckets:

• Motion ignored — the clip looks static or generic. Your reference likely had weak or ambiguous motion. Swap in a clip with one obvious, sustained movement.
• Reference copied too literally — you get the same scene with minor changes. Your "replace" instructions were too soft. State the new scene, wardrobe, and lighting as explicit changes.
• Muddy result — usually a busy reference or too many simultaneous changes. Simplify to one inherited motion plus one or two replacements, then build up.

Change one variable per retry: fix the reference first, then the prompt, then the settings. Changing all three at once tells you nothing about which one caused the improvement.

A quick decision check

Use reference-to-video when you have a clip whose movement or look is the point and you want new content in that mold. Use image-to-video when a single frame defines the shot and you only need it to move. Use text-to-video when you have no source material and want the model to invent the motion. And if your goal is to rework an existing clip rather than generate a new one from it, the Video editing option is the closer match.