Back to blog
Guides

How to Write NSFW Image-to-Video Motion Prompts

Write clearer NSFW image-to-video prompts with one subject action, controlled camera movement, pace, secondary motion, and stability instructions.

An image-to-video prompt should describe what changes after the first frame. The reference already supplies the subject, starting pose, composition, environment, and much of the visual style. Repeating every visible detail can bury the motion instruction.

The most useful prompt is not necessarily the longest. It gives the model one primary action, one camera behavior, a pace, limited secondary motion, and a clear stability priority.

Start with one subject action

Choose the movement that defines the clip. Keep it physically compatible with the starting pose and short enough for the selected duration.

Useful motion language is observable:

  • slowly turns toward the camera;
  • shifts weight and settles into a new pose;
  • raises one hand in a controlled movement;
  • breathes naturally while maintaining the pose;
  • makes a small change in expression.

Avoid stacking several actions with “then” unless the clip is long enough and the transitions are essential. A list of actions can compete for the same frames.

Add one camera instruction

Decide whether the camera moves or stays fixed. Examples include:

  • locked camera;
  • slow push-in;
  • gentle pull-back;
  • slight pan to the right;
  • restrained handheld movement.

Large camera movement changes the visible background and can introduce more drift. When subject consistency is the priority, start with a locked camera or a slow push-in.

Set the pace

Words such as slow, restrained, gradual, steady, or energetic change the temporal expectation. Match the pace to the duration. A five-second clip can show one compact action clearly; it is not a complete multi-shot sequence.

If the model rushes the action, reduce the number of requested changes before adding more adjectives.

Keep secondary motion subordinate

After the primary movement is clear, add one or two small environmental details when they support the scene:

  • subtle hair movement;
  • gentle fabric response;
  • soft light shift;
  • minor background depth movement.

Secondary motion should not compete with the subject. If the first test is unstable, remove it and establish the main action first.

State what should remain stable

Use a short stability instruction tied to visible features. Examples include maintaining facial identity, preserving the starting composition, keeping clothing details consistent, or holding background geometry steady.

Do not turn the final sentence into a long list of every possible defect. Prioritize the one or two elements that matter most for the clip.

Use a compact prompt formula

Try this structure:

[primary subject movement]. [camera behavior] at a [pace] pace. [one secondary motion]. Maintain [most important stable elements].

Example for a close portrait:

The adult subject slowly turns toward the camera and makes a small change in expression. The camera remains locked. Hair moves subtly while the background stays still. Maintain facial identity and the original lighting.

Example for a medium shot:

The adult subject makes one controlled pose adjustment and settles naturally. The camera performs a slow push-in. Fabric responds gently to the movement. Preserve body proportions, clothing details, and background geometry.

These are structural examples, not guaranteed recipes. The source frame and selected model still affect the result.

Match the prompt to the first frame

Read the prompt while looking at the source. Remove any instruction that contradicts the visible pose or crop. If the prompt asks for a full-body movement but the frame ends at the waist, create a wider source first.

Use the first-frame checklist to confirm that the source contains the space and detail needed for the action.

Choose duration after the action is defined

MuseUnbound currently supports 5-15 second image-to-video clips. Pick the shortest duration that can complete the action without rushing. Longer duration should serve a specific need, not compensate for an unclear prompt.

The current interface offers 768P and 1080P. Solve motion direction at a suitable test setting before treating higher resolution as a correction for prompt problems. Credit cost is displayed before generation and varies with the selected model and duration.

Run controlled prompt tests

Keep the reference, model, duration, and resolution fixed. Change one phrase between attempts:

  • subject action;
  • camera behavior;
  • pace;
  • stability priority.

Record what changed and what happened. If several variables change at once, the test cannot show which instruction improved or damaged the result.

Diagnose common prompt failures

If too little happens, make the primary action more observable rather than adding several new actions.

If too much changes, remove scene description and secondary motion. Keep the subject action and camera instruction.

If the camera moves in the wrong direction, place the camera behavior in its own sentence and remove conflicting cinematic language.

If identity or anatomy drifts, reduce the action, shorten the duration, and review the source image. Continue with the face, limb, and background drift guide.

Final prompt checklist

Before generating, confirm that the prompt:

  • describes one primary action;
  • contains no conflicting sequence of movements;
  • names one camera behavior;
  • sets a pace appropriate for the duration;
  • limits secondary motion;
  • identifies the most important stable elements;
  • matches the visible crop and pose;
  • complies with adult-only consent and content rules.

When the prompt passes this check, open the MuseUnbound image-to-video generator, select the duration and resolution, review the displayed credit estimate, and evaluate the whole output rather than only the first frame.