Wan 2.7 Video

Overview

wan2.7-video/image-to-video is an image-driven video generation model in the Alibaba WAN video series, accessed through the unified HiAPI.ai interface. It generates video starting from an image and also supports guidance from reference images/reference clips, allowing the visual content and camera movement to be controlled via prompts. Output resolution can be 720P or 1080P, each generation lasts from 2 to 15 seconds, and it supports combined inputs of first frame, last frame, driving audio, and an opening clip. It is suitable for developers who need to produce short videos from existing image assets, such as for advertising creatives, e-commerce showcases, and storyboard previews.

Key Features

• Image-driven generation with first/last frame control — You can specify both the first-frame and last-frame images to constrain the starting and ending composition of the shot, reducing regeneration costs caused by randomness.
• Reference clips and driving audio participate in generation — In addition to first and last frames, you can also pass in an opening clip and driving audio, giving the generated result a basis in motion rhythm and audio-visual alignment.
• Adjustable duration and resolution — Each generation can be set to last between 2 and 15 seconds, and the output resolution can be 720P or 1080P, adapting to different delivery requirements.

Capabilities

Image-to-video generation: Uses one or more images as the visual basis and generates video together with a text prompt of up to 5000 characters, making it suitable for turning existing visual assets into dynamic content.

Combined first-frame and last-frame control: You can specify both the first and last frames in a single request, and can also add an opening clip as the starting reference to define the starting and ending images of the shot.

Driving audio input: Supports passing in driving audio as a media input, so the generated video aligns in rhythm with the given audio material, reducing the need for later audio syncing.

Prompt extension and negative prompting: prompt_extend enables automatic prompt expansion, negative_prompt lets you declare visual elements to avoid, and seed allows reproducing the same result.

Parameterized resolution and duration: Output resolution and duration are both request parameters, so 720P/1080P and clip lengths of 2 to 15 seconds can be switched by scenario within the same interface.

Use Cases

• E-commerce and brands: generating dynamic showcases from product images — Input a product hero image and display requirements to produce a several-second product motion clip for detail pages, feed advertising, and social media assets.
• Content teams: turning character or scene images into shot clips — Use a character design image as the first frame or reference to generate short shots that match the storyboard description, for short films, animation previews, and creative pitches.
• Design teams: transition shots constrained by first and last frames — Provide both the starting and ending images to generate a transition clip between them, for title sequences, transitions, and visual concept presentations.
• Audio material owners: audio-driven visual generation — Pass in audio material and a visual description to generate video segments that match the rhythm of the sound, for short music videos and voice-over content.

Limitations and Selection Advice

The maximum duration per generation is 15 seconds; longer narratives require segmented generation and post-production splicing, and cross-segment consistency depends on prompt and reference material control. Output resolution is available in two tiers, 720P and 1080P; higher quality requirements need separate evaluation. Media inputs are limited to three items, and each type can appear only once. If the task is mainly text-based ideation and does not require an image basis, consider a text-to-video model from the same series; when ultra-long continuous shots are needed, this model is not a suitable choice.