LESSON 1 · GETTING STARTED

What WAN 2.7 Spicy and image-to-video are

WAN 2.7 Spicy is a WAN 2.7 image-to-video variant with extended support for explicit content. You provide one starting frame (an image) and a text prompt — the output is a single continuous shot that comes to life from that frame. The #1 beginner mistake: describing the picture instead of the motion.

How Spicy differs from base WAN 2.7
  • One input — only [Image 1]. No last-frame, video, audio, or extra references.
  • One scene — no montage or shot changes; the API blocks camera cut / multi-shot by default.
  • explicit — built for intimate motion; you need the motion vocabulary and strong Constraints.
  • Flexible duration — any whole number of seconds from 2 to 15.

Three ways to make an AI video

To avoid confusion going forward, let's sort out three neighboring concepts. They look alike, but their input and logic are different.

Text-to-video

The input is text only. The model invents both the scene and the motion on its own. Maximum freedom, minimum control over "how the frame looks." This is a separate t2v tool, not our mode.

Image-to-video (our mode)

The input is a single starting frame [Image 1] plus text. The look of the scene is already set by the picture, and the text controls what moves and how. Maximum control.

"Animating a picture"

A casual name for the same image-to-video, but often without a clear prompt — "just bring it to life." The result is random. We'll do the same thing, but deliberately and controllably.

An important difference from Kling and Seedance
WAN 2.7 Spicy has no "first + last frame" mode (interpolation between two pictures). WAN animates only FORWARD from the first frame. If what you need is a transition between two given frames — that's a different tool.

The core idea of the whole course

Remember this before any parameters or techniques. The model already sees the starting frame. You don't need to tell it what's depicted there — it already knows. That's why your entire prompt should describe the MOTION that begins from this frame: what the subject does, where and how the camera moves, the pacing, how the light changes, how the motion starts and how it settles.

Rule number one
Don't describe the picture — describe the motion. Re-describing what's already visible in the frame ("a calm scene, a woman with brown hair") is mistake #1 in image-to-video. Such a prompt gives the model not a single command, so it moves the frame at random.

Bad and good on the same example

Imagine a starting frame: a woman by a window, soft morning light. Here are two prompts for the very same frame.

A peaceful scene. A woman with brown hair stands by a window. Soft morning light. She looks calm and beautiful.

There isn't a single verb of motion here — just a retelling of the picture. The model doesn't understand what should happen.

The woman slowly turns her head toward the camera and softly smiles. Her hair drifts a little in the breeze from the window. The camera makes a gentle slow push-in. Calm, steady pacing; the motion settles quietly at the end of the shot.

The same frame, but now the subject's action, the camera movement, the pacing, and how the motion settles are all specified. This is exactly what the model will bring to life.

What a frame is made of in your head

Keep three roles in mind. Each is responsible for its own part of the result — and you must not mix them up.

STARTING FRAME

What is shot

The image [Image 1] sets the look: the character, the composition, the light, the style. This is "what's already in the frame." Without a starting frame, image-to-video won't run at all.

PROMPT

What happens

The text sets the motion: the subject's action, the camera movement, the pacing, the light changes, the motion arc. This is "what will happen starting from the frame."

PARAMETERS

How to render

Duration, resolution, seed, negative_prompt. Technical settings on top of the frame and the prompt.

Key WAN 2.7 Spicy parameters

We'll go through each of them in detail in the following lessons, but it's useful to know the basic boundaries right now.

ParameterValuesIn brief
Resolution720P / 1080P / 1080P-SR / 1440P-SR480p not supported. SR tiers upscale from 720P.
Duration2–15 sec (any integer)Need it longer — stitch clips by the last frame.
Inputonly imageOne first frame. No last-frame, video, audio, or references.
SeedseedRepeatability. -1 — random.
Negative promptnegative_promptAPI adds a shot-change baseline + your artifact list.
1
starting frame as input
15s
maximum for a single clip
1440P
maximum (SR tier)
Why Spicy
The Spicy WAN 2.7 variant holds one continuous shot, stable motion physics, and supports explicit content — as long as you describe motion with verbs and set strong Constraints (anatomy, limbs, skin).

Course map

Next we'll build up control over motion step by step. Here's the route.

01
Getting started
this lesson: the model sees the frame, we describe the motion
02
Anatomy of a prompt
subject + action, camera, pacing, light, motion arc
03
Camera movement
pan, dolly, push-in, orbit, crane and how to align them with the action
04
Pacing and arc
calm → peak → settle, "throughout the shot"
05
Parameters and negative
resolution, duration 2–15s, negative_prompt
10
Spicy & explicit
motion vocabulary, Constraints, template
09
Long videos
stitching clips by the last frame beyond 15 seconds
image-to-video starting frame motion prompt WAN 2.7 Spicy describe the motion

Try it yourself in ZenCreator

Generate your first clip — everything this course teaches is available in the ZenCreator video tools.

Start Creating
Next lesson