Choose the WAN3 workflow that matches the material you already have

01

Three production paths in one video model

Use text-to-video for a written brief, image-to-video with a first frame or first-and-last frame pair, or reference-to-video when visual, motion or audio references need to guide the result.

• Text-to-video starts from a prompt without source media.
• Image-to-video accepts one first frame or a first-and-last frame pair.
• Reference-to-video accepts one to five mixed image, video or audio references.
• Audio output, seed and watermark controls appear only where the current WAN3 contract supports them.

02

Output controls and duration limits

WAN3 produces 720p or 1080p video. Text and frame workflows accept an explicit 2–30 second output. Reference workflows use the same range for image or audio references; when a video is included, its rounded-up duration plus the selected output must stay within 30 seconds.

• Prompt limits and supported advanced controls are validated before submission.
• A reference request accepts at most one video alongside other supported references.
• Every duration and resolution shown in the Workspace is priced before submission.
• Every successful output is transferred to WAN3-managed storage before the task is marked complete.

03

Know the credit quote before generation

WAN3 costs 25 credits per billable second at 720p or 53 credits per billable second at 1080p. Standard tasks have a two-second minimum; the Workspace calculates and displays the authoritative quote before submission.

04

Plan around source and delivery limits

Source files must match the formats and size limits shown in the composer. Tasks are private to the account owner, and queued, processing, succeeded, failed and refunded states remain visible in Creations with an owner-only download when the result is ready.