VideoAlibaba

Wan 2.7

Wan 2.7 is Alibaba's video model, notable for taking a reference image together with a first frame and for exposing a seed and a negative prompt. Reach for it when you want repeatable results rather than a new roll every time.

Duration2 to 10s

Resolutions720p, 1080p

Aspect ratios1:1, 4:3, 3:4, 16:9, 9:16

AudioNone

Reference inputsUp to 4

Make with Wan 2.7 →View pricing

Added July 2026. Included on every plan.

Why this one

What makes Wan 2.7 different

A seed you controlLock the seed and a prompt returns the same clip, which turns iteration into editing rather than gambling.
A negative promptSay what you do not want. Few video models here accept that, and it is the fastest way to kill a recurring artifact.
Reference plus first frame togetherMost models make you choose. This one takes a reference alongside the opening still.
Short clips are welcomeThe duration range starts lower than most, which suits loops and cutaways.
Five aspect ratiosSquare through wide and vertical, covering the usual placements.
Picture onlyNo generated audio to strip before the clip reaches an edit.

Use cases

Best for

Repeatable variations
Loops and cutaways
Product turns
Style matching
A/B tests
Background plates

Technique

Prompt tips

Fix the seed once it looks closeThen change one word at a time. This is the model where that workflow actually works.
Use the negative prompt for artifactsName the thing that keeps appearing: extra limbs, text, watermarks, blur.
Combine the reference with the first frameIdentity from the reference, composition from the frame. That pairing is this model's advantage.
Start shortThe low end of the duration range is cheap and tells you quickly whether the motion idea works.
Describe motion plainlyDirection and speed beat adjectives on this model.

Mechanism

How Wan 2.7 works

Wan takes a prompt, an optional first and last frame, and an optional reference image, and renders with a seed you can pin. Because the seed is exposed, the same inputs reproduce the same clip, which is what makes the negative prompt useful: you can remove one unwanted element and see only that change rather than an entirely new take.

Getting started

How to use it on Dolly

Pick Wan 2.7Choose it in the video studio, or just ask the agent for it by name.
Write the promptDescribe what you want, and add your reference images or frames where they help.
GenerateThe exact credit cost sits on the button before you press it, and a failed run refunds automatically.

Pricing

What it costs

Every generation shows its exact credit cost on the button before you run it.

720p · 5s: 78 credits
1080p · 5s: 117 credits

One run is about 1/16 of a Starter month.

Failed generations refund automatically. See plans

Capabilities

Specs

Resolutions: 720p, 1080p
Aspect ratios: 1:1, 4:3, 3:4, 16:9, 9:16
Durations: 2 to 10s
Frame inputs: first + last
Reference images: Up to 4
Audio: None
Settings: Seed, Negative prompt
Max per run: 2

Questions

Wan 2.7 FAQ

What does the seed actually do?

It fixes the randomness. Same prompt and same seed gives the same clip, so you can iterate deliberately.

How long can the negative prompt be?

Up to 500 characters, which is enough for a list of artifacts and not enough for a second brief.

Can I use a reference and frames together?

A reference works alongside the first frame. That combination is the reason to pick this model.

Does it generate sound?

No, picture only, which leaves a clean plate for an edit that brings its own audio.

How many references can I give it?

Up to 4 per run, and a reference works alongside the first frame rather than instead of it.

Ready to make with Wan 2.7?

Every model is included on every plan. The price sits on the button before you run it.

Start making →