Wan 3.0 Video

Overview

wan3.0-video-prime is the high-quality reference-to-video line in the Wan 3.0 Video family, accessible through the unified HiAPI.ai interface. It accepts reference materials such as images, audio, documents, and web pages, and combines them with prompts to generate videos with sound up to 30 seconds long and up to 1080P, with support for first/last frame control, duration, and aspect ratio selection. It is suited to developers and content teams that need to maintain subject consistency using brand or product materials and produce finished clips directly.

Key Highlights

• Reference-driven multimodal generation — a single call can combine up to 10 images, 5 audio clips, and document or web page references, cited in the prompt as Image1, Audio1, and so on in array order.
• Single-call duration from 2 to 30 seconds — the duration parameter directly determines output length, so long narratives can be generated as a complete clip in one pass without segment stitching.
• Up to 1080P with selectable tiers — three resolution tiers (480P, 720P, 1080P) and six aspect ratios let you trade off quality against cost per call.
• Synchronized audio track and first/last frame control — output includes synchronized audio by default and the audio track can be disabled; first-frame and last-frame images constrain the opening and closing shots.

Capabilities

Reference-consistent generation: lock in subjects and style with reference materials such as images and audio, keeping the same character or product visually consistent across the entire video and reducing post-production rework.

Structured material input: pass in one document or one login-free web page as a reference, and use the prompt to turn the material into explainer-style video visuals.

Controllable duration and aspect ratio: choose any duration from 2 to 30 seconds and any of six aspect ratios; adaptive automatically follows the input materials and prompt, making it easy to fit landscape, portrait, and other placements.

First/last frame shot constraints: specify first-frame and last-frame images separately to constrain where a shot starts and ends, suited to finished clips that need precise openings and closings.

Reproducibility and audio toggle: reproduce the same result with a fixed seed, and use the audio parameter to control whether a synchronized audio track is generated, which helps with batch generation and version comparison.

Use Cases

• Brand or product materials into ad clips — input product images and selling-point prompts to output short clips with sound that keep the product's appearance consistent, for e-commerce detail pages and social media placements.
• Multimodal reference character narrative segments — use character images plus audio references to generate coherent clips up to 30 seconds as material segments for storylines or short dramas.
• Materials into explainer videos — upload a document or web page link to generate explainer visuals on the corresponding topic, for internal training and product introductions.
• Shot completion with specified start and end frames — provide two keyframe images for the first and last frames to generate a transition shot of controllable length and feed it into post-production editing workflows.

Limitations and Selection Advice

Billing is based on output seconds; higher resolution and longer duration cost more, so long videos combined with high quality should be budgeted in advance. Generation quality depends heavily on the prompt, and results vary considerably when the subject, action, camera, and style descriptions are unclear. Reference materials and first/last frame modes cannot be mixed; a single call accepts only one document or one web page reference. For shorter, lighter everyday clips, choose the standard line in the same family; for batch tasks with strict requirements on fixed parameters and reproducibility, fix the seed before scaling up.