Wan 2.5 Image-to-Video

About Wan 2.5 Image to Video

Wan 2.5 image to video starts from one still and one prompt and returns a five- or ten-second clip with an audio track. The image fixes the subject, framing and light; the prompt supplies the action, camera move and pacing. Every other field in the panel is optional: an audio upload, a negative prompt, prompt expansion and a seed.

The three resolutions are priced in five-second blocks and the figure updates before you submit. At 480p a five-second run costs 300 credits, at 720p it costs 600, and at 1080p it costs 900; choosing ten seconds doubles each number. The endpoint's own wait estimate climbs the same ladder, at roughly 240, 360 and 420 seconds. That spread is wide enough that the practical question is not whether to finish at 1080p but whether the shot reads at 720p or even 480p first, since a draft pass at the lowest tier costs a third of the top one.

Audio is what separates this page from most image-to-video tools on the site. A narration or music bed up to 40MB can be attached, and the model is asked to sync lip movement or lock the motion to the beat; leave the field empty and it generates a soundtrack itself. A negative prompt takes a comma-separated list of what should stay out of frame, prompt expansion lets the model enrich a short prompt with its own detail, and a fixed seed makes a take you liked repeatable. The reference image accepts JPEG, PNG, WebP or GIF up to 20MB, either dragged in or pasted as a hosted URL, and the prompt is the one field that cannot be left blank.

Against the fast page, this is the slower and more expensive path that keeps the 480p tier. Against Wan Animate and Wan 2.2 Animate, the distinction is the input: both of those take a driving clip and restage a character inside it, while Wan 2.5 image to video has no footage to follow and has to infer all motion from the text. If the shot needs a specific performance, this is the wrong page. If it needs a clean still brought to life with sound, it is the shorter route.

How Wan 2.5 Image to Video works

1. 1

   Add the still and describe the shot

   Drag in a JPEG, PNG, WebP or GIF up to 20MB, or paste a hosted image URL. The prompt is required: name the action, the camera move, the style and the beat you want, in that order.

2. 2

   Set length, resolution and sound

   Choose five or ten seconds, then 480p, 720p or 1080p; the credit cost above the button recalculates at 300, 600 or 900 per five seconds. Optionally attach audio up to 40MB as MP3, WAV, M4A, OGG or WebM.

3. 3

   Tune, generate, download

   Add a negative prompt for artifacts you keep seeing, switch prompt expansion on for terse prompts, and fix the seed to repeat a run. A failed job returns its credits, and the finished MP4 keeps its audio when downloaded.

What Wan 2.5 Image to Video is best for

• Product stills that need motion

  A clean packshot becomes a short moving shot; the 1080p tier covers storefront playback and the 480p draft confirms the camera move before you pay for it.

• Portrait plus your own narration

  Upload the voice track alongside the image and the model is asked to sync the mouth movement to it, which is faster than rebuilding the timing by hand.

• Ten-second inserts

  The duration switch covers the length most edits actually need, at double the five-second price rather than a separate model with its own limits.

• Reproducible iterations

  Keep the seed, image and prompt fixed and change one variable at a time; with prompt expansion off, the only thing moving between runs is what you edited.

Wan 2.5 Image to Video at a glance

Model

Alibaba Wan 2.5 image to video, run through WaveSpeed

Clip length

5 or 10 seconds

Resolution

480p, 720p or 1080p

Credit cost

300, 600 or 900 credits per five seconds by resolution; a ten-second run doubles it

Inputs

Reference image up to 20MB (JPEG, PNG, WebP, GIF) or a hosted URL plus a required prompt; optional audio up to 40MB, negative prompt and seed

Output

MP4 with an audio track, downloaded from the preview panel

Wan 2.5 Image to Video FAQ

How much does a Wan 2.5 image to video run cost?

300 credits per five seconds at 480p, 600 at 720p and 900 at 1080p, doubled for a ten-second clip. The estimate sits above the generate button and is charged once per request; a job that returns no video gets the credits back. New accounts open with 300 credits and can claim 50 per daily login, both expiring after 30 days, and the starter pack is 500 credits for $9.90.

What happens if I leave the audio field empty?

The model produces its own soundtrack, so the clip is never silent unless you ask for silence in the prompt. Uploading a track instead changes the job: the model is asked to match mouth movement to the speech or the motion to the beat, which is what makes a narration or a music bed land on time.

Should I use this page or the fast one?

This page keeps the 480p tier at 300 credits per five seconds, which is the cheapest way to test a shot. The fast endpoint drops 480p, charges 420 credits at 720p and 630 at 1080p, and returns sooner. Both accept the same image, prompt, audio, negative prompt, seed and the same 5 or 10 second durations.

Can I get the same result twice?

Set a seed and keep everything else identical: the same image, the same prompt, the same resolution and duration. Prompt expansion adds text of its own, so turn it off if you want a strict repeat. A seed only carries meaning inside this endpoint; it does not transfer to the fast variant.

Can it copy motion from a clip I already have?

No. This page has one image and text, so every movement in the output is inferred. If the performance already exists on video and you need it preserved, use Wan Animate or Wan 2.2 Animate, which take the driving footage and a character image as their two required inputs.