Image-to-Video Prompt Structure: How to Animate Still Images with AI

Image-to-video (img2vid) is fundamentally different from text-to-video prompting. The input image carries all the visual information — your job is to tell the model how to move it, not what it looks like. We tested this at HackAIGC across hundreds of NSFW stills, and the results were clear: most people over-prompt and break their source image.

The Fundamental Rule of Img2Vid

Don't describe what's in the image. Describe what the image does.

The model sees your input. Every word you use to describe the subject creates an opportunity for contradiction with the source image, which causes morphing and distortion. Keep subject descriptions out of your prompt entirely.

The Img2Vid Prompt Framework

[Motion direction], [Animation type], [Focus area],
[Tempo/pace], [Camera instruction], [Duration hint], [Anti-distortion quality tags]

Motion Direction

The single most important element in any img2vid prompt. Your source image is static — direction is what brings it to life.

Direction	Prompt Tag	Best For
Outward	`motion from center outward, expanding movement`	Reveal shots, stretching
Upward	`upward motion, rising action, vertical flow`	Standing up, raising arms
Horizontal	`subtle side-to-side sway, gentle horizontal motion`	Dancing, relaxing, standing
Inward	`inward motion toward center, converging movement`	Hugging, intimacy, focus
Depth	`forward motion toward camera, depth traveling`	Walking toward viewer, interaction

Animation Type

Choose based on the composition of your source image.

Type	Prompt Tags	Works Best With
Breathing	`gentle breathing animation, subtle life added, chest rise and fall, natural micro-motion`	Portraits, lying poses
Parallax	`cinematic parallax effect, layered depth movement, background separates from subject`	Full-body, scenery shots
Full body	`full body animation, complete figure motion, natural body movement across frame`	Standing poses, active scenes
Micro-motion	`micro-expressions, slight head tilt, subtle eye movement, minimal facial motion`	Close-up faces

Focus Area

The input image determines the subject, so tell the model where to animate.

Focus	Prompt Tags	Best Image Type
Face	`focus motion on face area, micro-expressions, eye movement, slight head turn`	Portrait, close-up
Hair/Clothing	`hair and clothing animated, flowing fabric, wind effect, fabric movement with body`	Full body, outdoor
Full figure	`full body animation, complete figure motion, natural body movement`	Standing, active pose
Background only	`background motion only, subject remains still, environmental animation`	Wide shot, object focus

The NSFW Img2Vid Anti-Distortion Stack

The biggest problem with NSFW img2vid is that aggressive motion distorts anatomy. These tags preserve quality:

preserve subject structure, maintain original anatomy,
no body distortion, keep face consistent,
stable figure throughout, no morphing,
image fidelity preserved, prevent anatomy warp

Why this matters: NSFW content has lower error tolerance for distortion. A distorted face in a landscape shot is acceptable; a distorted body in an NSFW clip ruins the entire output. We never skip this stack.

Prompt Build Examples

Example 1: Portrait Animation (NSFW Close-up)

Input: Close-up portrait of a woman looking at camera

gentle breathing animation, subtle life added to static image,
focus motion on face area, micro-expressions, eye movement,
slight head turn, very gentle motion,
subtle hair movement, warm atmosphere, slow tempo,
preserve subject structure, maintain original anatomy,
butter-smooth animation, 5 seconds, cinematic quality

Example 2: Full Body Reveal

Input: Standing full-body shot of a woman in lingerie

full body animation, complete figure motion across frame,
natural weight shift between legs, subtle hip sway,
hair and clothing animated, gentle flowing fabric,
slow pan up from feet to head, vertical camera movement,
preserve subject structure, maintain original anatomy, stable figure,
smooth transition, 5 seconds, ultra HD quality

Example 3: Bedroom Intimacy

Input: Woman lying on bed, side view

gentle breathing animation, subtle chest rise and fall,
micro-movements in hands and fingers, slight head tilt,
hair spread on pillow animated with slow movement,
background motion only, soft lighting shift,
gentle zoom in toward face, warm atmosphere,
preserve subject structure, no body distortion, stable figure,
slow tempo, 6 seconds, cinematic

Img2Vid vs Text-to-Video: Cheat Sheet

Aspect	Image-to-Video	Text-to-Video
Subject description	❌ Skip entirely	✅ Required (4+ identifiers)
Motion direction	✅ Essential	✅ Important
Camera movement	⚠️ Optional (can conflict)	✅ Important
Quality tags	✅ Keep minimal	✅ Required
Style tags	❌ Image already has it	✅ Required
Duration	✅ Include	✅ Include

FAQ

Q: Why does my img2vid result look distorted?

A: Almost always because you described the subject in the prompt, creating a contradiction with the source image. Remove all subject descriptions and focus only on motion and camera instructions.

Q: How much motion is safe for img2vid?

A: Less than you'd think. The source image has its own visual structure — aggressive motion breaks it. Gentle, slow motion preserves image quality. We recommend starting with `gentle` and `subtle` as your default motion modifiers for NSFW content.

Q: What's the best duration for img2vid?

A: 4-6 seconds. Longer durations cause the model to drift from the original image. For longer clips, generate multiple segments.

Q: Does image resolution matter?

A: Yes. Higher input resolution produces smoother animation. 1024x1024 minimum. Lower resolutions produce jittery or distorted motion, especially with active movement.

Q: Can I animate non-human NSFW content?

A: Yes. For object-focused content, use environmental animation: lighting shifts, subtle background movement, gentle camera push. Don't try to animate the object itself — animate around it.

Try HackAIGC's Uncensored Image Generator | Image Generator | Video Generator