01
Plan a complete shot
Text and frame workflows support an explicit 2–30 second result, leaving room for an opening, action, camera move and ending beat.
WAN3 AI video generator and Wan series studio
Direct longer AI video with WAN3.
Create 2–30 second AI video from text, one or two frames, or a focused reference set. Compare WAN3 with Wan 2.5, 2.6 and 2.7 before choosing the workflow that fits your material.
Draft here. Review the exact workflow and credit quote in Workspace.
What is WAN3?
The long-form, multi-input path in the Wan series.
WAN3 is the lead AI video model in this studio for shots that need more time or more source direction. It supports text-led creation, one or two guiding frames, and reference-led generation without exposing internal delivery providers.
- maximum selected output
- 30s
- highest current resolution
- 1080p
- reference slots
- 5
02
Start from the material you have
Write from scratch, guide the result with one or two frames, or combine one to five focused image, video or audio references.
03
Keep sound in the same brief
Generate audio with the video or use audio references in the WAN3 reference workflow when sound direction matters.
04
Know the quote before submission
WAN3 is quoted at 25 credits per billable second for 720p or 53 credits per second for 1080p, with a two-second minimum.
WAN3 vs the Wan series
Choose by the job, not the version number.
WAN3 is the longest current output path. Wan 2.7 offers the broadest editing toolkit, Wan 2.6 adds fixed video-to-video work, and Wan 2.5 keeps text and single-image creation simple.
WAN3, Wan 2.7, Wan 2.6 and Wan 2.5 compared
The live Workspace remains authoritative, but this table gives a fast pre-generation decision based on the current public capability and pricing contract.
| Decision | WAN3 | Wan 2.7 | Wan 2.6 | Wan 2.5 |
|---|---|---|---|---|
| Best fit | Longer prompt, frame or reference-led shots | Reference control and directed video editing | Fixed text, image or video-to-video tasks | Straightforward text or single-image clips |
| Input paths | Text, 1–2 frames, or 1–5 mixed references | Text, frames, one clip, references or video edit | Text, one image, or 1–3 videos | Text or one image |
| Duration | 2–30s; video reference + output must fit 30s | Workflow-specific 2–15s | 5/10/15s; video-to-video 5/10s | 5 or 10s |
| Resolution | 720p / 1080p | 720p / 1080p | 720p / 1080p | 720p / 1080p |
| Credit path | 25 credits/s at 720p; 53 credits/s at 1080p | 20 credits/s at 720p; 30 credits/s at 1080p | Fixed quotes from 120 to 525 credits | Fixed quotes from 100 to 335 credits |
Three WAN3 creation paths
Begin with text, frames or references.
Pick the workflow from the source material you can actually provide. Detailed controls, validation, cost and formal submission stay inside your private workspace.
01
Text to video
Describe the subject, action, camera, light, sound and timing for an explicit 2–30 second shot.
02
Frame-guided video
Use one image as the opening direction or add a second frame to define where the shot should land.
03
Reference to video
Combine up to five purposeful image, video or audio references; if video is included, its rounded duration plus output must fit within 30 seconds.
How to use WAN3
Build the shot in four deliberate passes.
01
Write the shot arc
State the opening image, subject movement, camera path, atmosphere and final beat instead of relying on generic quality words.
02
Choose only useful references
Give every frame, clip or audio file one job—identity, environment, motion, sound or the intended ending—and remove conflicting material.
03
Match output to delivery
Select duration, 720p or 1080p resolution and the aspect ratio required by the final placement before reviewing the quote.
04
Review continuity, then iterate
Check subject identity, camera direction, motion and the ending. Revise the exact instruction or reference that missed rather than rewriting the whole brief.
WAN3 model specifications
The current public contract, in one place.
These are the controls WAN3 can validate before submission today. The Workspace hides unsupported combinations and remains the final source for each task quote.
- Workflows
- Text to video, frame-guided video, reference to video
- Selected output length
- 2–30 seconds
- Resolution
- 720p or 1080p
- Frames
- One required image, optional second image
- Reference workflow
- 1–5 total references; no more than one video
- Aspect ratios
- 16:9, 4:3, 1:1, 3:4, 9:16
- Advanced controls
- Seed and generated audio
- Pricing
- 20 credits/s at 720p; 30 credits/s at 1080p; 2-second minimum
WAN3 prompt examples
Write a shot plan, not a pile of adjectives.
Use these structures as starting points. They describe subject, movement, camera, light, sound and an ending without pretending that an example is a verified generated result.
Prompt 0101
Product launch film
A matte black running shoe rests on wet stone at night. Slow dolly-in as orange light travels across the mesh, then cut to a low orbit while water beads lift from the sole. End on a clean side profile with the laces settling. Deep room tone, one restrained impact at the final frame.
Prompt 0202
Cinematic story beat
At blue hour, a cyclist stops beneath an empty flyover as rain begins. Handheld camera follows from behind, moves alongside when the streetlights switch on, then rises into a wide overhead view as the cyclist enters the glowing city. Tire spray, distant traffic and a low synth pulse.
Prompt 0303
Vertical campaign cut
9:16. A silver jacket moves through warm fog in a dark studio. Begin with a close texture detail, pull back as the model turns toward camera, then let three controlled light sweeps reveal the silhouette. Finish on a still front-facing pose with space above for campaign copy.
What creators can build
Use the extra time for a clearer visual idea.
A longer output is most useful when the scene has a reason to develop. WAN3 fits product, narrative and social work that benefits from a planned opening, movement and ending.
01
Product launches and ads
Move from a clean product reveal into material details, controlled camera motion and a final brand-ready hero frame.
02
Story beats and pre-visualization
Test how a scene establishes location, develops one action and resolves before committing to a larger production plan.
03
Vertical and social campaigns
Plan a 9:16 fashion, travel or campaign cut with intentional pacing instead of stretching a short motion loop.
Credits without surprises
See the real estimate before the task exists.
Your exact quote depends on the selected model, workflow and supported settings. The workspace calculates it from the current server contract.
WAN3 / Credits
- The homepage never charges credits.
- The workspace confirms the current server-calculated estimate.
- A covered provider submission failure follows the generation refund path.
WAN3 questions
Choose the model with the boundaries visible.
What is WAN3 best used for?+
WAN3 is best for 2–30 second text, frame or reference-led shots that need a planned opening, movement and ending. It supports 720p and 1080p output, generated audio and up to five references in the reference workflow.
Should I choose WAN3 or Wan 2.7?+
Choose WAN3 when you need a longer selected output of up to 30 seconds and its three focused creation paths. Choose Wan 2.7 when you need the wider editing toolkit, including source-clip, reference and directed video-edit workflows.
How many references can WAN3 use?+
The reference workflow accepts one to five total image, video or audio references, with no more than one video. When a video reference is present, its rounded-up duration plus the selected output must stay within 30 seconds.
How are WAN3 credits calculated?+
WAN3 costs 25 credits per billable second at 720p or 53 credits per billable second at 1080p, with a two-second minimum. The Workspace shows the authoritative server quote before submission.
Does the homepage start a generation?+
No. It saves a private, time-limited draft and opens the Workspace. Only the Workspace Generate action creates a task and applies credits.
Which other models are available?+
The same private Workspace also includes Wan 2.5, Wan 2.6, Wan 2.7 and Seedance for video, plus Nano Banana and GPT Image for image creation. Each model exposes only its current supported controls.
Ready when the shot is
Build the first take, not the final guess.
Carry a private direction into Wan Studio and make the production choices where they belong.