Xyphrax Updated Advanced Image Instructions

These new Advanced Image Instructions add extra information for Camera, Orientation, Back to Camera Shots, better positioning.

It is suggested to use a tracker for appearance and clothes with these instructions.

They are longer and will cost a few more credits but the outcome is amazing.

This a a combined work put together from aspects and collaboration with KaapstadMK, MrDrunkinDragon, Thyr

These instructions are to be used without a Image Style in the World, you will add the Image style in the instructions themselves. This will keep the style even if the word length does go over.

You can also update the direction.

Screenshot on how Image Style should Look using these instructions:

CRITICAL CONSTRAINTS:
1. Prioritize high-weight visual keywords. Avoid fluff words like "beautiful", "stunning", or "hyper-detailed".
2. Do not include any introductory or concluding text. Output ONLY the raw prompt.

GLOBAL RULES:
- Use comma-separated visual descriptors.
- Avoid complete sentences.
- One to three-word descriptors preferred.
- Total prompt ≤200 words (hard cap 250).
- No pronouns — refer to subjects by ID only.
- Eye color: one word (e.g. "Amber") never describe a large or expressive.
- NEVER use forms of: "blush," "flush," "nipples," "areolas."
- No clothing or expressions inside illustrAppearance (separate fields).


SUBJECT SELECTION:
- illustrSubject prioritize two subjects, identifiable characters by ID.
- Always use a first and last name for subjects in this format (first_last)
- No identifiable character in input → use (player_name).
- Subject interacts with player → include (player_name) as one subject.
- NEVER "Crowd," "Object," or "Environment" as subject. 

ORIENTATION RULES:
- Prioritize positional logic over facial expression every time.
- Use the back facing camera illustrAppearance rule whenever the scene describes a character as Fighting , facing away, Intimate, or Dramatic Scenes with Back to Camera.
- For scenes involving interaction, conversation, or intimacy: Subjects MUST face each other in profile or 3/4 profile. 
- NEVER use the phrase "facing the camera" or "angled toward the viewer" if the subjects are interacting.
- NEVER use "Not looking at Camera" as this reinforces looking at the camera
- Only have subjects face the camera if it is a solo portrait or a specific voyeuristic moment where they are posing for the viewer.
- Intimate encounters should be dramatic and with descriptive interaction, sometimes this means they have their back-to-camera (no facial features) facing away from camera.
- Couch subject facing in the scene, facing objects/other subjects. NEVER mention the camera.

COMPOSITION:
- Clearly state subject placement.
- Examples: centered, left foreground, right foreground, overlapping silhouettes, profile-to-profile, back-to-camera foreground.
- Positional composition is more important than facial expression.

CINEMATOGRAPHY:
- Directed as "ENTER YOUR OWN DIRECTION"
- Drone-style direction — full range of angles (high/low/close-up/wide).
- Fully rendered lighting (skin reflection, glow on stone, face lighting, shadows).
- Fights → wide-angle, Combat scenes must depict the moment of impact or immediately before impact. Weapon trajectories and spell effects must clearly show cause and target.

CAMERA SELECTION:
Choose the single best-suited camera, lens, and aperture for the scene from the options below. Use only the quoted text in output.
- Camera: "Shot on Sony A7IV" · "Canon 5D Mark IV" · "Hasselblad X2D" · "Fujifilm X-T5"
- Lens: "14-24mm" (wide, dramatic) · "35-50mm" (natural, versatile) · "70-85mm" (portrait, flattering) · "100mm+" (telephoto, background separation)
- Aperture: "f/1.4-f/2.8" (shallow DOF, blurred BG) · "f/4-f/5.6" (moderate, balanced) · "f/8-f/16" (deep focus, everything sharp)


IMAGE STYLE (apply consistently):
- Base: "YOUR IMAGE STYLE" always add at the end of the image style "IWUpscaleFaceSmooth" *REMOVE IWUPSCALE IF ANIME*

VARIABLE SCHEMA:

illustrSubject: (turn_number):(<id_1>)(<id_2>)

illustrAppearance: IMAGE STYLE: [artistic medium / color palette / lighting / detail / shading / brightness / contrast] (N) people shown:
• Facing camera / normal interaction: (<id_1>): [ethnicity, sex, height, build, hair, eye color, skin tone, desirable physical traits, non-human features if any, grime/dirt/blood]
• Fighting / facing away / Intimate or Dramatic Scenes with Back to Camera: (<id_1>): [back-to-camera, no facial features — sex, height, build, hair, skin tone, grime/dirt/blood]
• Multi-subject: (<id_1>): [...] | (<id_2>): [...]

illustrClothes: (<id_1>): [garments, fabric, material, fit on body, wear and damage, exposed areas from damage and quality of that exposure] | (<id_2>): [...]

illustrExpressionPosition: DIRECTOR INFO: [Camera, Lens, Aperture, camera angle, exposure, artistic direction, coloring, tone] | (<id_1>): COMPOSITION: [orientation, subject placement, where looking, pose, action, emotion, interaction with other subjects] | (<id_2>): [...]


illustrClothesChanged: (<id>): [TRUE]
illustrAppearanceChanges: (<id>): [NONE]

illustrSetting (3 layers, characters FORBIDDEN here):
- Layer 1 (background): distant environment — materials, textures, lighting, shadows, time of day.
- Layer 2 (midground): immediate environment — materials, textures, lighting, shadows. Crowds belong here.
- Layer 3 (foreground): weather effects, objects in front of subjects. 

FINAL OUTPUT (every field filled):
illustrSubject: ...
illustrAppearance: ...
illustrClothes: ...
illustrExpressionPosition: ...
illustrClothesChanged: ...
illustrAppearanceChanges: ...
illustrSetting:
  - Layer 1 (background): ...
  - Layer 2 (midground): ...
  - Layer 3 (foreground): ...

FINAL OUTPUT RULES:
Word limit (hard cap): Single subject: 200 words total. Dual: 100 words per subject. COUNT before outputting. If over: TRUNCATE immediately — setting first, clothing behavior second, standard appearance third. Never truncate non-human or transformation features. Exceeding corrupts image output and is never acceptable.
Field length (hard cap): illustrAppearance - 50 words maximum; illustrClothing - 50 words maximum; illustrExpressionPosition - 50 words maximum; illustrSetting - 50 words maximum.