How many scenes should an AI story have?

It depends on the platform. TikTok stories work best with 8-15 scenes (30-60 second video). Website stories can be 15-30 scenes. A full graphic novel chapter might use 20-40 scenes. Start with fewer scenes and add more only if the story needs them.

How do I make scenes look connected in AI art?

Use three anchors: (1) identical character description prefix in every prompt, (2) consistent art style keywords, (3) environment transitions that make visual sense - go from wide to medium to close-up rather than jumping randomly between locations.

What is the best camera progression for AI stories?

Start with a wide establishing shot to set the scene. Move to medium shots for dialogue and interaction. Use close-ups for emotional moments. Return to wide shots when changing locations. This mimics how actual films are shot and feels natural to viewers.

How do I show emotions in AI-generated characters?

Add specific expression keywords to your prompt: 'furrowed brows, clenched jaw, narrowed eyes' for anger. 'Eyes welling with tears, downturned lips, slumped shoulders' for sadness. 'Wide grin, crinkled eyes, raised cheeks' for joy. Physical body language matters too.

Can I use the same seed to keep scenes consistent?

Using the same seed helps with consistency but limits composition variety. A better approach is to use a consistent character prefix plus --cref or IP-Adapter for face consistency, while allowing the seed to vary so each scene has unique composition and camera angles.

Scene-by-Scene AI Prompting: How to Build a Full Visual Story

What Is Scene-by-Scene Prompting?

Scene-by-scene prompting is the technique of writing a series of connected AI image prompts that tell a coherent visual story. Each prompt builds on the previous one - sharing the same character, progressing the environment, and advancing the emotional arc. It is the difference between random AI images and a professional visual narrative.

Think of each prompt as a screenplay direction for a single frame. When done well, the resulting images flow together like stills from a film. This guide teaches you exactly how to write these connected prompts, based on the techniques we use for our own AI visual stories.

Planning Your Scene List

How Many Scenes?

The number of scenes depends on your format and story complexity. A simple TikTok story needs 8-12 scenes. A web-based story works well with 15-25. A full graphic novel chapter might use 30-40. Start with the minimum scenes needed to tell the story - you can always add more.

Scene Pacing

Not every scene should be the same intensity. Great stories alternate between three types of beats:

Action beats: Something happens - movement, conflict, discovery. Wide or dynamic camera angles. 2-3 seconds per image when converted to video
Emotional beats: Character reacts - close-ups on face, expression keywords, intimate lighting. 3-4 seconds to let emotion land
Transition beats: Scene changes - establishing shots of new locations, time-of-day shifts, environmental storytelling. 2 seconds to orient the viewer

Establishing shot to close-up sequence showing progressive camera work

Writing Scene Prompts That Connect

Identical Character Descriptions

The character description prefix must be identical in every single prompt. Copy and paste it - do not retype it or paraphrase it. Even small changes like "brown hair" vs "dark brown hair" can cause the AI to generate different-looking characters. For full character consistency techniques, see our character consistency guide.

Environment Transitions

When moving between locations, make the transition feel natural. Do not jump from a cozy bedroom to a dark forest with no connection. Instead, show the character leaving (medium shot at doorway), traveling (wide shot of path), and arriving (establishing shot of new location). Three transition scenes make any location change smooth.

Time-of-Day Progression

Use lighting to show time passing. Morning scenes use "soft golden sunrise light, cool blue shadows." Midday uses "overhead bright sunlight, sharp shadows." Evening uses "warm golden hour, long shadows, amber tones." Night uses "moonlight, cool blue tones, artificial light sources." Consistent time-of-day changes make your story world feel real.

Emotional Arc Through Expressions

Expression progression showing four different emotions on the same character

Your character's emotional state should evolve across scenes. Map out the emotional journey: curious > excited > worried > terrified > relieved. Each scene prompt changes only the expression and body language keywords while keeping everything else constant.

Expression keywords to master:

Joy: wide genuine smile, crinkled eyes, raised cheeks, relaxed posture
Fear: wide eyes, raised eyebrows, mouth slightly open, tense shoulders, leaning back
Anger: furrowed brows, clenched jaw, narrowed eyes, rigid posture, fists clenched
Sadness: downcast eyes, slightly pursed lips, slumped shoulders, head tilted down
Surprise: raised eyebrows, wide open eyes, open mouth, leaning forward

Prompt Template per Scene Type

Annotated scene breakdown with labeled prompt components

Establishing Shot

[Character prefix], standing at the entrance of [location], looking out at [environment details], wide establishing shot, [time-of-day lighting], [atmosphere keywords], [art style], 8K, cinematic composition

Dialogue and Interaction

[Character prefix], [action - talking, gesturing, looking at something], [expression], medium shot from waist up, [indoor/outdoor environment], [lighting], [art style], detailed facial expression, 8K

Action Scene

[Character prefix], [dynamic action - running, fighting, reaching], [intense expression], dynamic low angle shot, motion blur on background, [dramatic lighting], [environment], [art style], high energy, cinematic, 8K

Emotional Close-Up

[Character prefix], extreme close-up on face, [specific expression with micro-details], [emotional lighting - warm for joy, cool for sadness], shallow depth of field, blurred background, [art style], ultra detailed, 8K

Climax and Reveal

[Character prefix], [climactic action or reaction], [peak expression], dramatic low angle, [most intense lighting of the story], [dramatic environment], [art style], epic cinematic composition, volumetric lighting, 8K

Real Example: Breaking Down a 20-Scene Story

Here is how a complete story breaks down using these techniques. This is the structure we used for our story "Pushed Overboard":

Scenes 1-3 (Setup): Wide establishing shots of the location, medium shots introducing the characters, warm friendly lighting. Expressions: relaxed, happy.
Scenes 4-6 (Rising tension): Medium shots showing growing conflict, lighting becomes more dramatic, expressions shift to concerned and suspicious. Camera angles tighten.
Scenes 7-9 (Escalation): Dynamic angles, dramatic lighting, intense expressions. Action beats mixed with emotional close-ups.
Scenes 10-12 (Climax): Most dramatic camera angles, peak lighting contrast, extreme expressions. This is where all the tension pays off.
Scenes 13-15 (Aftermath): Camera pulls back to wider shots, lighting softens but stays tense, expressions show shock and processing.
Scenes 16-20 (Resolution): Gradual return to calmer composition, lighting reflects the new emotional state, expressions show transformation.

Free Scene Prompt Templates

Download our character prefix template and scene prompt templates from the character consistency guide and combine them with the scene type templates above. The formula is always: Character Prefix + Action + Expression + Environment + Camera + Lighting + Style + Quality.