Thyr Image Instructions

Thyr Image Generation Instruction Templates

• If you have any questions or want to find my worlds, you can search "_Thyr_" on both Discord and IW.

Many of the Templates below work way better with the Main Cast" and "Player Information" tracked items. Find them here:
https://infiniteworlds.mywikis.wiki/wiki/Thyr_Tracked_Item_Templates

NOTICE: Cheaper AI models may not be able to use all of the features of some templates. All of these templates were created utilizing Smilodon and Smilodon-thinking.

Up-Front Credits: Thank you to @Xyphrax for the base template and feedback and collaborative work while I was making Lite and Full. Your enthusiasm about the “BodyOnly_” process and the “Camera/Lens/Aperture Selection" Improvements were great. @WolfishGrimm and @Obscure Ideolect for play testing and giving great feedback. And @KaapstadMK for rolling in some of my ideas into their IGI!

Turn-number process and my Recommendation(s)

For my templates I always force a turn number to be generated. I did this mainly because when I first used Xyphrax's template I just thought it was a nice feature. didn't really realize it forced a re-writing of fields because every turn's illustrSubject line was unique. This does, of course- lead to some immersion-breaking moments where characters' appearances change dramatically because the AI didn't retain the information well, etc. Thyr Lite can fully function without turns- just delete the bolded section under illustrSetting.

So any templates that have a turn number rely heavily on either the Summarization instructions (in the specialist instruction area) explicitly telling the AI to retain X information per character (and it succeeding), or with a tracked item (or other method) of retaining appearance information for the image generation.

My personal preference is to have a tracked item that the image generation instructions references. And within the rules for that tracked item- I allow the AI to update everything except for the characters appearance. I always play in storyteller mode so I'm able to go in and make updates as needed to an appearance if there's a permanent change which is not a big deal for me. But some people don't like this and that's perfectly OK. I am very nitpicky about immersion in my games and one of my biggest pet peeves is when a character randomly becomes a new race a couple hundred turns later.

So overall- if you are going to use one of my instructions , simply copy and paste it into an extra instruction block. Then I would recommend setting up a tracked item. I do have a couple of templates linked above you can start with- all of my instructions reference a tracked item named main cast and with the three section version it splits between Main Tracker and Main cast. I personally recommend that you have any of these trackers include any information you want from your important characters. This should be current clothes, inventory, appearance settings which I would lock to only be edited by you, their name, and then anything else pertinent to your story.

To be frank I just like to test new ways of doing image instructions. One may work better for you and another might not. All of them should utilize the same tracked item version so you can plug and play and test them out as you wish.

My current favorite is the 3-step. Simpler. READ ABOUT IT CAREFULLY BELOW!

Variable Settings:

First- all of my templates follow these rules for where information goes within the variables.

• illustrIsCharacter is always True.
• illustrSubject is always (firstname_lastname) for each character. Usually I enforce turn numbers to force all variable prompts to be reset and rely on tracked item to ensure main character appearance consistency (and other things) which is most affected by AI's memory loss. so its always (Turn_Number): (first_last1)(first_last2)
• illustrAppearance is only the base description of a character, hair worn down, Ancestry (Main Inst. cross-table if non-earth world to use earth ethnicities in image gen only) Also because of front prompts "weighing" more, I usually plop [Director info] | [Thematic Instructions] | in front, then: [# of illustrsubjects] | [Character 1 name and desc, then 2, etc.]. Intimate body part descriptions are strictly forbidden (only in illustrClothes when visible) here, as well as ANY temporary effects. Appearance stays basic and clean.
• illustrClothes is where intimate body descriptions go (otherwise if a character wears concealing clothing but the illustrApperance gives detailed descriptions of whats under the clothes, the image will alter the clothes to make those parts visible). So- if illustrClothesChanged is set to True after the outcomeDescription list a change- these descriptions naturally would need to occur at the same time. and of course clothes go here. :p
• IllustrExpressionPosition is now also where any temporary effects go. Magic, hair worn a certain style, wounds that are visible, etc. Also expressions and positions.
• illustrSetting is the same- but using a two-level layer system for background/mid and foreground, as opposed to Xyphrax's original 3 he had on his.
• The other variables do not change

Different Instruction Templates:

Each Template is designed a bit different. Some work best with an accompanying Tracked Item- also make sure your “Specialist Instructions” for Summarization specifically says to track character appearances in detail. Copy and paste the instructions for the one you chose into an extra instruction block. Further below there are also some instructions for a tracked item- highly recommended.

Thyr Lite IGI:

Variable Restrictions for image clarity, Director Info, Thematic Instruction, “BodyOnly_” generation for when characters are facing away and their faces and other features/clothes are not visible, ~600 word instruction block. Light weight and should get the job done.

Thyr Full IGI:

Started with Xyphrax’s template and incorporated Variable Restrictions, then created the “BodyOnly_” generation for when characters are facing away and their faces and other features/clothes are not visible. Includes Director Info with Camera/Lens/Aperture Settings, Image Thematic Guidance (built from Main Instruction and outcomeDescription analysis to give establishing Theme and directive information) ~1200 word instructions. Works best with Tracked item or Keyword Instruction Block to house Main Cast Appearances and Current Clothes/Main outfit (Armor or equipment depending on world).

Thyr 12-Step IGI:

Same bells and whistles as the Thyr Full IGI. Also works best with a tracked item or keyword instruction block. The only difference is layout- It uses a Step-by-step process. Best for newer users to the IGI process to understand. Same output as the Thyr Full IGI. More words

Thyr 3-Section IGI:

This is my current favorite image generation instructions because it is the least amount of words, but has all of the features that you would find in the other image generation instructions , but also includes image guidance so that you can leave your image prompts completely blank and you can override with basic tags and it also generates the image theme and it places it at the beginning of the entire prompt in the illustration appearance field because that is where that information is most weighted. It's broken into three sections. First is Critical Rules overall. Then the Features narrated separately. Followed by the Output section where it lists for each variable the restrictions and then the output format.
Here's how to set it up:

1. Copy and paste the template into a extra instruction block from below.
2. Make a tracked item (Check out my templates from above link or make your own. regardless I recommend that you put in the update instructions the actual template and then all of the changeable information within brackets. Then you can explicitly tell the AI they are only allowed to update information within the brackets. But I would also recommend restricting them from being able to update the appearance information that would later populate into the illustration appearance field.
3. Ensure that the two locations in the instructions that are bolded are filled in with the name(s) of your tracked item(s) so the AI can find it.
4. Change the image settings of your images and delete all of the information. This is the area where you would select the preset like dark fantasy or other preset. For this simply leave it empty.
5. Update the Bolded and Italicized IMAGE STYLE OVERRIDE section with what you want your image to look like. recommend keeping the tags in all caps and separated by commas.

Templates:

Thyr Lite Template

<Image Generation Instructions>

**PERSISTANT DATA SOURCES**
Always check Tracked Items for persistent character appearance or current clothes tracking. If information is found for an illustrSubject's Appearance or Clothing- prioritize what is found in tracked items over narrative information and adjust the narrative to align. [Enter your tracked item or Keyword instruction block name into this section if you have one.]

**THEME FILTERS**
Gather data from world information, main instructions, and outcomeDescription to describe the tone, genre, theme, environment, emotional feel, etc. and narrate in less than 30 words. Output [THEME FILTERS output]

**OUTPUT AND RESTRICTIONS**
**illustrSubject:** (Must be up to three identifiable characters in last outcomeDescription. If none- set as <<player_name>>. No Pronouns. (Firstname_Lastname) format only) Output: '(Turn_Number):(<id_1>)(<id_2>)'
**illustrIsCharacter** [True]
**illustrAppearance:** (If concealing clothing is worn- remove body type and bust size descriptions. Temporary appearance descriptors of nipples, genitals, aereolas, blush, wounds, temporary effects, and skin abnormalities from mood or lack of sleep/food are forbidden here.) Output: `(Insert this text: "IWUpscaleFaceSmooth") | (number of illustrSubject) PEOPLE SHOWN: (<id>): [Earth ethnicity/sex/apparent age/height/body type/hair color and style/eye color(single color or hyphenated only)/bust size(ex. "DD-cup bust")/skin tone/non-human feature/non-human feature ] | (<id>): [...]`
**illustrClothes:** (If naked- only put "Completely Naked-" and describe nipples, genitals, other intimate details.) Output: `(<id>): [Nipple and genital description if visible, Clothing item 1- fabric/fit/color/design/detail, Clothing item 2 - fabric/fit/color/design/detail, ...] | (<id>): [...]`
**illustrExpressionPosition:** (characters or body parts of non-subjects forbidden from Director Info.) Output: 'DIRECTOR INFO: [The Artistic direction of the shot, camera angle, skin reflection, glow on stone, face lighting, shadows] | (<id>): [Pose/Action/Emotion/Temporary magical visual effects/visible wounds/interaction with other illustrSubject] | (<id>): [...]`
**illustrClothesChanged:** (if one of the illustrSubject changes clothes in the latest outcomeDescription- change Output [Data] to [True]. Otherwise leave as [False]) Output: `(<id>): [Data] | (<id>): [Data]`
**illustrAppearanceChanges:** (If outcomeDescription lists illustrSubject having PERMANENT change to eye color, hair color or style, body type, bust size, skin appearance or color, or a non-human feature- then summarize new appearance in [Data] for each character.) Output: `(<id>): [Data] | (<id>): [Data]`
**illustrSetting:** (This variable MUST contain two layers. Characters/Subjects are forbidden here. Move non-subject NPCs/Crowds to Layer 1.) Output: '[THEME FILTERS output] (Generate a sentence describing the genre, atmosphere, theme, etc. of the shot based on the main instructions and outcomeDescription for this turn)
- LAYER 1 (background):
    - Environment: [visual details only, distant environment, materials, textures, lighting, shadows, time of day]
- LAYER 2 (midground and foreground overlays):
    - Environment: [immediate environment, materials, textures, lighting, shadows, weather effects, objects in front of illustrSubject]'

**BODYONLY_ID GENERATION:** CRITICAL: BodyOnly variants are turn-specific rendering instructions ONLY. They exist exclusively in illustration fields for the current turn. They NEVER persist in Main Cast, Character Relationship Tracker, Overall Summary, Current Situation, Plot Threads, Future Notes, secretInfo, or any narrative text.

If any (<id>) faces away from <<Player_Name>> or their face is not visible due to position/angle/hair:

Create temporary rendering variant: (BodyOnly_Firstname_Lastname)
Modify illustrAppearance: Retain only visible elements (skin tone, body type excluding bust unless visible from angle, visible non-human features like tails/wings). Format: "[Gender] Body, Head fully obscured by [hair style and color] hair, No face visible, [DO NOT GENERATE FACIAL FEATURES]". If bald: "face obscured by back of bald head with [skin color] skin."
Modify illustrClothes: describe only visible clothing elements
Complete remaining steps for BodyOnly variant
Use BodyOnly variant in illustration fields ONLY for this turn
Next turn: Discard BodyOnly variant entirely, begin fresh from base (<id>) entry
BodyOnly variants evaporate at turn end. 

When summarizing events involving BodyOnly-rendered characters, reference ONLY base character name:

CORRECT: "T51: <<Player_Name>> spoke with Yuki_Shirakawa at the railing"
FORBIDDEN: "T51: <<Player_Name>> spoke with BodyOnly_Yuki_Shirakawa at the railing" 

The illustration used a BodyOnly variant, but the event involved the base (<id>). Summaries track narrative events, not rendering techniques.

</Image Generation Instructions>

Thyr Full Template

<Image Generation Instructions>

**STEP 1. VARIABLE DEFINITIONS**

1A: **illustrSubject:** `(<id_1>)(<id_2>)`
1B: **illustrAppearance:** `(<id>): [Ethnicity/Sex/Apparent Age/Height/Build/Hair color, style, and cut/Eye/Breast size(Cup size and ONLY describe with "Bust", like "DD-Cup Bust")/Skin tone/Desirable features (Remove descriptors of their figure when wearing baggy clothing and NEVER include temporary appearance descriptors like being tired, wounded, being dirty, etc. This should reflect them in a normal state. descriptions of nipples, genitals, and aereolas are forbidden here.)] | (<id>): [...]`
1C: **illustrClothes: `(<id>):** [Describe each article of clothing with Fabric and Design/Material/Wear/How clothing fits on body, If nipples or genitals are exposed, describe here in detail alongside worn clothing. If Naked, put "Completely Naked"] | (<id>): [...]`
1D: **illustrExpressionPosition:** `(<id>): [Pose/Action/Emotion/temporary magical effects/wounds that are not concealed by clothes/Specific interactions only with a illustrSubject that is in the scene] | (<id>): [...]`
1E: **illustrClothesChanged:** `(<id>): [TRUE] ONLY if any of the 'illustrSubject' changed their clothes or outfit in the 'outcomeDescription', otherwise [FALSE] by default.`
1F: **illustrAppearanceChanges:** `(<id>): [Short description of the new specific changes in appearance based on the latest outcomeDescription. Must be an attribute defined in 1B. like Hair, eyes, etc.Concise as possible]`
1G: **illustrSetting:** (This variable MUST contain two layers. Characters/Subjects are forbidden here. Move non-subject NPCs/Crowds to Layer 1.)
- Layer 1 (background):
 - Environment: [visual details only, distant environment, materials, textures, lighting, shadows, time of day]

- Layer 2 (midground and foreground overlays):
 - Environment: [immediate environment, materials, textures, lighting, shadows, weather effects, objects in front of illustrSubject]

**STEP 2. RULES AND CONSTRAINTS**

2A: **NO PRONOUNS:** Use IDs (e.g., player, subject name) exclusively. MUST BE (First name_last name) FORMAT.
2B: **EYE COLOR:** One word only (e.g., "Green"), or hyphenated for dual color (e.g., "Grey-Orange").
2C: **IDENTIFIABLE CHARACTERS:** The `illustrSubject` variable MUST only be exactly 1, 2, or 3 specific, identifiable characters. 
2D: **THE FALLBACK RULE:** If no specific, identifiable characters are described in the input text, set the `illustrSubject` to `<player_name>`.
2E: **CROWD/OBJECT FILTER:** NEVER set a "Crowd," "Object," or "Environment" as the subject. If a crowd exists but is not the subject, move it to `illustrSetting [Layer 1]`.
2F: **DIRECTOR INFO:** 

Select camera, lens, and aperture best suited for scene:

-Camera Options: "Shot on Sony A7IV" or "Canon 5D Mark IV" or "Hasselblad X2D" or "Fujifilm X-T5"

-Lens Selection: "14-24mm" - Wide angle, dramatic perspective; or "35-50mm" - Natural, versatile framing; or "70-85mm" - Portrait, flattering compression; or "100mm+" - Telephoto, strong background separation

Aperture Control: (1.4 - to 16 is a RANGE. any number that falls between is acceptible. Like: f/1.8) Range options: "f/1.4 - f/2.8" - Shallow depth, blurred background, or "f/4 - f/5.6" - Moderate depth, balanced sharpness, or "f/8 - f/16" - Deep focus, everything sharp.

Then add artistic direction: camera angle, skin reflection, lighting on surfaces, face lighting, shadows, atmospheric effects - separated by commas.

Final format:
DIRECTOR INFO: [Camera, lens, aperture, artistic elements]

2G: **IMAGE THEMATIC GUIDANCE:**

Generate 7 tags to establish compositional and emotional tone for image generation:

*3 tags from Main Instructions:*
-Extract thematic keywords from world setting (post-apocalyptic, maritime survival, factional conflict)
-Extract tonal keywords from narrative style (gritty, dark fantasy, gallows humor)
-Extract genre identifiers (adult adventure, survival, political intrigue)

*4 tags from current outcomeDescription:**
-Identify primary emotional tone of scene (tense, intimate, violent, contemplative)
-Identify narrative moment type (confrontation, quiet character moment, action sequence)
-Identify environmental atmosphere (cramped tavern, windswept docks, underground chamber)
-Identify character relationship dynamic present (hostile, protective, sexual, wary) 

Synthesis Process:
Combine 7 tags into single concise sentence establishing image composition guidance. Focus on: genre classification, emotional tone, compositional mood, narrative weight, atmospheric quality.

Example outputs:
-"Dark post-apocalyptic maritime survival scene with intimate vulnerability, gritty realism, and protective tension in cramped quarters"
-"Brutal factional confrontation with gallows humor undertones, dangerous power dynamics, windswept coastal atmosphere, and wary distrust"
-"Quiet character development moment in survival setting, emerging intimacy against harsh world, contemplative mood with underlying danger"

This guidance informs overall image composition without dictating specific visual elements—establishing emotional and tonal framework for AI interpretation.

2H: **TRACKED ITEMS:** By default, when a illustrSubject ID has been selected for image generation, all tracked items and keyword instruction blocks should be checked for character appearance or clothing descriptions. If a tracked item or keyword instruction block is found with relevant information, it must be prioritized over other sources of information. The player can also input the name of their tracked item or keyword instruction block below, meaning that step 2G will skip the search function and simply go to specified Tracked Item or keyword instruction block.

Character Description Tracked Item Override(s):
[]

2I: **BODYONLY_ID GENERATION:** For any 'illustrSubject's, there may be scenes where the/multiple 'illustrSubject' is present, but the scene requires their face to be obscured and not visible, any mention facing away, back facing <<player_name>>, face looking down or obscured, etc. Triggers this step. Some examples may be if the shot is of them from behind, if they are engaging is sexual activity where their face is not visible, or they are completing a task and their head is down, etc. In this circumstance- as the absolute last step of the image generation process, first determine what parts of the appearance and clothing should be visible based on their positioning, then follow these steps only for the applicable (<id>):

- Step 1: Change the illustrSubject's (<id>) to include "Bodyonly_" before their name. like: (Bodyonly_John_Doe). Make this change in all prompts.
- Step 2: Change 'illustrAppearance'- Remove all descriptors and only retain descriptors of skin, and their build (exclude bust unless visible), state their gender followed by "Body" (like "Female Body"), followed by inserting "Head fully obscured by (Hair style and color) hair (unless bald, then just enter "face obscured by back of bald head with X skin color"), No face visible, [DO NOT GENERATE FACIAL FEATURES]". 
- Step 3: Change 'illustrClothes'- Remove descriptive elements that describe features of their outfit that would not be visible. If the image is a shot of their back, any descriptors of the front of their shirt, jacket, pants, jewelry, or other items should be removed, and the back of their outfit described. If naked, only put "Naked-" and describe nipples, aereolas, tattoos and genitals only if visible.
- Step 4: Change their 'illustrExpressionPosition'- list their position, parts of the body visible facing the viewer/camera, and what surface they are on devoid of emotions, descriptors of appearance, clothes, etc. if genitals are visible state it. for this section keep it as brief as possible with few words.
- Step 5: repeat this process for other 'illustrSubject' if applicable.

2J: **AFFECTING THE NARRATIVE RULE**: These RULES Are NOT story related rules. This is only applied to what is written and sent to the Image generator. This will have no impact on the story and cannot be used to cue moments. 
 
**STEP 3. STRICT OUTPUT TEMPLATE**

illustrSubject: (<<Turn_Number>>):(<id_1>)(<id_2>)
illustrAppearance: [IMAGE THEMATIC GUIDANCE OUTPUT] | DIRECTOR INFO: [Data as concise natural description] | (Insert this text: "IWUpscaleFaceSmooth") | (number of subjects) PEOPLE SHOWN | (<id_1>): [Data] | (<id_2>): [Data]
illustrClothes: (<id_1>): [Data] | (<id_2>): [Data]
illustrExpressionPosition: (<id_1>): [Data] | (<id_2>): [Data]
illustrClothesChanged: (<id_1>): [Data] | (<id_2>): [Data]
illustrAppearanceChanges: (<id_1>): [Data] | (<id_2>): [Data] | [...]
illustrSetting:

- Layer 1 (background):
 - Environment: [Data]

- Layer 2 (midground and foreground overlays):
 - Environment: [Data]

</Image Generation Instructions>

Thyr 12-Step Template

<Image Generation Instruction Steps>

* **FOUNDATIONAL PRINCIPLES:** *
All image generation must serve narrative consistency and visual storytelling. Character appearance data resides in Main Cast tracker as canonical source. BodyOnly variants exist only for turn-specific rendering and never persist in any form.

* **NOTATION SYSTEM:** *
Characters selected as illustrSubject are formatted as (Firstname_Lastname). Throughout these instructions, this will be abbreviated as (<id>). When multiple subjects are present, they increment as (<id_1>), (<id_2>), (<id_3>).

* **PERSISTANT DATA SOURCES** *
Alway check Tracked Items for persistant character appearance or current clothes tracking. If information is found for illustrSubject Appearance or Clothing- prioritize what is in tracked items over narrative information and adjust the narrative to align. [Or enter the name of your tracked item of keyword instruction block with character information here]

**MANDATORY EXECUTION ORDER:**

**STEP 1: POV VISIBILITY VERIFICATION**
-Before proceeding with any image generation, answer: Can <<Player_Name>> see all (<id>)s' faces from current position and angle based on outcomeDescription?
-If NO for any (<id>) → Proceed immediately to STEP 10 (BodyOnly Generation) for affected (<id>)s.
-If <<Player_Name>> is alone and walking/traveling/brooding with back to viewer → Proceed to STEP 10 for <<Player_Name>> as sole (<id>).
-Common outcomeDescription flags requiring BodyOnly: "bent forward," "turned their back," "facing away," "looking over shoulder," "ass toward," "walked ahead."

**STEP 2: CHARACTER IDENTIFICATION**

-Check Main Cast tracker for (<id>) appearance data. If (<id>) exists in Main Cast, use that data as canonical source. If not present, generate new (<id>) following standard format.
-For first-time (<id>) generation, format as:
	-(<id>): [Ethnicity/Sex/Apparent age/Height/Body type/Hair color and style/Eye color (format: "human [color] eye color with dark pupils")/Bust size if applicable (format: "[X]-cup bust")/Skin tone/Non-human features (tails format: "[color] [type] tail from tailbone")]
-At this step: illustrIsCharacter = TRUE, illustrClothesChanged = FALSE, illustrAppearanceChanges = NONE

**STEP 3: SUBJECT SELECTION**

-Select 1-3 interacting characters from outcomeDescription as illustrSubject in (Firstname_Lastname) format, referenced as (<id_1>), (<id_2>), (<id_3>) when multiple.
-Priority: Active participants in current scene > Characters being addressed > Environmental elements only if no characters present.
-If no character interactions occur, set illustrSubject to <<Player_Name>> only.
-If meta conversation with AI occurs, set illustrSubject as (<<Player_Name>>)(Friendly_Fox) and format as <<Player_Name>> talking to friendly-looking fox on luxurious chair.
-If <<Player_Name>> looks into mirror, set them as illustrSubject.

**STEP 4: CLOTHING ANALYSIS**

-Compare outcomeDescription clothing descriptions to existing illustrClothes in Main Cast.
-If outfits differ for any (<id>), set illustrClothesChanged = TRUE and describe new outfit: (<id>): [Outfit description - materials, fit, condition, layers. If genitals/nipples/breasts/ass fully exposed, state explicitly with anatomical detail. If naked, write "Naked-" followed by intimate body description including exposed genitals. NEVER describe nipples or piercings as "visible through" fabric - only when completely exposed.]

**STEP 5: PERMANENT APPEARANCE CHANGES**

-Did (<id>)'s Ethnicity, Sex, Apparent Age, Height, Body Type, Hair color/style, Eye Color, Bust Size, Skin Color, or Non-Human features change PERMANENTLY in outcomeDescription?
-FORBIDDEN: Temporary visual effects (magic glows, environmental dirt, sweat, wounds, wet hair, removed jacket). Only permanent changes qualify.
-If permanent change detected, describe in illustrAppearanceChanges: (<id>): [Describe permanent change concisely without embellishment] | (<id_2>): [...]

**STEP 6: EXPRESSION AND POSITION**

-Following STEP 1 POV rules, describe visible facial expression and body position.
-Format in illustrExpressionPosition: (<id>): [Emotion/expression, position if not standing]
-Examples: "smiling, relaxed stance" / "concerned, kneeling" / "neutral, sitting"

**STEP 7: DIRECTOR INFORMATION**

Select camera, lens, and aperture best suited for scene:

-Camera Options:
"Shot on Sony A7IV" or
"Canon 5D Mark IV" or
"Hasselblad X2D" or
"Fujifilm X-T5"

-Lens Selection:
"14-24mm" - Wide angle, dramatic perspective; or
"35-50mm" - Natural, versatile framing; or
"70-85mm" - Portrait, flattering compression; or
"100mm+" - Telephoto, strong background separation

Aperture Control: (1.4 - to 16 is a RANGE. any number that falls between is acceptible. Like: f/1.8)
"f/1.4 - f/2.8" - Shallow depth, blurred background
"f/4 - f/5.6" - Moderate depth, balanced sharpness
"f/8 - f/16" - Deep focus, everything sharp

Then add artistic direction: camera angle, skin reflection, lighting on surfaces, face lighting, shadows, atmospheric effects - separated by commas.

Final format in illustrExpressionPosition:
DIRECTOR INFO: [Camera, lens, aperture, artistic elements] | (<id>): [Pose/Action/Emotion/Temporary magical effects/visible wounds] | (<id_2>): [...]

**STEP 8: SETTING DESCRIPTION**

NEVER list specific characters or NPCs in illustrSetting. Crowds permitted.

Format:
(Background): [Visual details, distant environment, materials, textures, lighting, shadows, time of day]

(Midground and Foreground): [Immediate environment, materials, textures, lighting, shadows, weather, objects in front]

Always state time of day ("daytime," "at night," "at dawn," "dusk").

**STEP 9: GENITALS AND EXPOSURE LOGIC**

IMPORTANT: Genital descriptions appear ONLY in illustrClothes field when explicitly exposed.

illustrAppearance contains permanent physical characteristics only - never temporary visibility states or genital descriptions.

If (<id>) is naked: set at "Naked" and describe their body and genitals, nipples, and other features in detail.

If (<id>) is clothed but nipples or genitals are exposed- describe both the clothing worn and the exposed genitals or breast/nipples.

FORBIDDEN: "nipples visible through," "outline visible through"

**STEP 10: BODYONLY GENERATION (EPHEMERAL RENDERING VARIANTS)**

CRITICAL: BodyOnly variants are turn-specific rendering instructions ONLY. They exist exclusively in illustration fields for the current turn. They NEVER persist in Main Cast, Character Relationship Tracker, Overall Summary, Current Situation, Plot Threads, Future Notes, secretInfo, or any narrative text.

If any (<id>) faces away from <<Player_Name>> or their face is not visible due to position/angle/hair:

Create temporary rendering variant: (BodyOnly_Firstname_Lastname)
Modify illustrAppearance: Retain only visible elements (skin tone, body type excluding bust unless visible from angle, visible non-human features like tails/wings). Format: "[Gender] Body, Head fully obscured by [hair style and color] hair, No face visible, [DO NOT GENERATE FACIAL FEATURES]". If bald: "face obscured by back of bald head with [skin color] skin."
Modify illustrClothes: Following STEP 1 POV rules, describe only visible clothing elements
Complete remaining steps for BodyOnly variant
Use BodyOnly variant in illustration fields ONLY for this turn
Next turn: Discard BodyOnly variant entirely, begin fresh from base (<id>) entry
BodyOnly variants evaporate at turn end. They never appear in:
	-Tracked Items
	-Overall Summary references
	-Current Situation descriptions
	-Plot Threads
	-Future Notes
	-secretInfo narrative
	-outcomeDescription text

When summarizing events involving BodyOnly-rendered characters, reference ONLY base character name:

CORRECT: "T51: <<Player_Name>> spoke with Yuki_Shirakawa at the railing"
FORBIDDEN: "T51: <<Player_Name>> spoke with BodyOnly_Yuki_Shirakawa at the railing"

The illustration used a BodyOnly variant, but the event involved the base (<id>). Summaries track narrative events, not rendering techniques.

**STEP 11: IMAGE THEMATIC GUIDANCE**

Generate 7 tags to establish compositional and emotional tone for image generation:

*3 tags from Main Instructions:*
-Extract thematic keywords from world setting (post-apocalyptic, maritime survival, factional conflict)
-Extract tonal keywords from narrative style (gritty, dark fantasy, gallows humor)
-Extract genre identifiers (adult adventure, survival, political intrigue)

*4 tags from current outcomeDescription:**
-Identify primary emotional tone of scene (tense, intimate, violent, contemplative)
-Identify narrative moment type (confrontation, quiet character moment, action sequence)
-Identify environmental atmosphere (cramped tavern, windswept docks, underground chamber)
-Identify character relationship dynamic present (hostile, protective, sexual, wary)

Synthesis Process:
Combine 7 tags into single concise sentence establishing image composition guidance. Focus on: genre  classification, emotional tone, compositional mood, narrative weight, atmospheric quality.

Example outputs:
-"Dark post-apocalyptic maritime survival scene with intimate vulnerability, gritty realism, and protective tension in cramped quarters"
-"Brutal factional confrontation with gallows humor undertones, dangerous power dynamics, windswept coastal atmosphere, and wary distrust"
-"Quiet character development moment in survival setting, emerging intimacy against harsh world, contemplative mood with underlying danger"

This guidance informs overall image composition without dictating specific visual elements—establishing emotional and tonal framework for AI interpretation.

STEP 11: FINAL OUTPUT COMPILATION

Convert all bracketed data to cohesive, concise natural narrative descriptions retaining all information. Multiple (<id>)s separated by " | ".

illustrSubject: (<id_1>)(<id_2>)
illustrAppearance: IWUpscaleFaceSmooth | [NUMBER] PEOPLE SHOWN | (<id_1>): [Data converted to concise natural narrative] | (<id_2>): [Data converted to concise natural narrative]
illustrClothes: (<id_1>): [Data as concise natural description] | (<id_2>): [Data as concise natural description]
illustrExpressionPosition: DIRECTOR INFO: [Data as concise natural description] | [IMAGE THEMATIC GUIDANCE OUTPUT] | (<id_1>): [Data as concise natural description] | (<id_2>): [Data as concise natural description]
illustrClothesChanged: (<id_1>): [TRUE/FALSE] | (<id_2>): [TRUE/FALSE]
illustrAppearanceChanges: (<id_1>): [Data or NONE] | (<id_2>): [Data or NONE]
illustrIsCharacter: TRUE
illustrSetting:
(Background): [Data as concise natural description]

(Midground and Foreground): [Data as concise natural description]

</Image Generation Instruction Steps>

Thyr 3-Section Template:

<illustrInstructions>

-----RULES:-----

**MAIN TRACKER**
Check [Main Tracker] and [Main Cast] each turn for character data. Tracker overrides narrative discrepancies.

**IMAGE STYLE OVERRIDE**
[PHOTO REALISTIC, CHIAROSCURO, DARK FANTASY, OIL PAINTING, WARM LIGHT]

**POV**
<<player_name>> is "the viewer". If <<player_name>> is not an illustrSubject or [BackgroundChar] in  illustrSetting, AL PROMPTS where the illustrSubject(s) is/are interacting with <<player_name>> must be phrased as interacting with/towards "the viewer" for a POV-like perspective.

**CRITICAL CONSTRAINTS**
-It is forbidden to have more than 2 illustrSubjects unless the player explicitly states it for that turn only.


-----FEATURES:-----

**BODYONLY GENERATION (EPHEMERAL RENDERING VARIANTS)**

CRITICAL: Turn-specific rendering ONLY. Exists exclusively in illustration fields. NEVER persists in Tracked  items, summaries, secretInfo, or narrative text.

TRIGGER: outcomeDescription states/implies (<id>)'s body and/or face is facing away from <<Player_Name>> (back turned, bent  away, presenting from behind, walking away, body oriented opposite direction) for any length of time.

IMPLEMENTATION:

Variant ID: (BodyOnly_Firstname_Lastname)
illustrAppearance: ONLY POV-visible elements (skin tone, body type, non-human features). Format: "[Gender] Body,  Head fully obscured by [hair style/color] hair, No face visible, [DO NOT GENERATE FACIAL FEATURES]" — If bald: "face obscured by back of bald head with [skin color] skin."
illustrClothes: ONLY POV-visible clothing
illustrExpressionPosition: "face not visible, [position]"
Discard entirely next turn. Resume base (<id>).
SUMMARY RULE: Reference base character name ONLY. Never reference BodyOnly variant in summaries.

**IMAGE THEMATIC GUIDANCE**
Extract from outcomeDescription and synthesize into ONE concise sentence:
World setting keywords + narrative tone + genre + emotional tone + moment type + atmosphere + relationship dynamic
Output: Single sentence establishing genre, emotional tone, compositional mood, narrative weight, atmospheric quality.

**DIRECTOR INFORMATION**
Select per scene:
Camera: Sony A7IV | Canon 5D Mark IV | Hasselblad X2D | Fujifilm X-T5
Lens: 14-24mm (wide/dramatic) | 35-50mm (natural) | 70-85mm (portrait) | 100mm+ (telephoto)
Aperture: f/1.4-2.8 (shallow) | f/4-5.6 (moderate) | f/8-16 (deep)
Artistic direction: camera angle, skin reflection, surface lighting, face lighting, shadows, atmosphere
Format: DIRECTOR INFO: [Camera, lens, aperture, artistic direction]

**IMAGE PROMPT CONSOLITATION**
-Consolidate [DIRECTOR INFO] and [THEMATIC GUIDANCE] into a cohesive, consice, and eloquent description. Output is [IMAGE GUIDANCE DATA].


-----OUTPUT:-----

Do not retain brackets [] or format outlines  in final output.

**illustrSubject:**
Restrictions: MAX 2 characters from outcomeDescription. Default <<Player_Name>> if none. (Firstname_Lastname)  format. Parentheses retained.
Format: (turn_number):(<id_1>)(<id_2>) TURN NUMBER IS CRITICAL

**illustrIsCharacter:** true

**illustrAppearance:**
Restrictions: BASE descriptions ONLY. No temporary effects. Data format: [ETHNICITY in CAPS, sex, apparent age,  height, body type, hair color+style, eye color (single), bust size (Small-Extremely massive), skin tone, non-human if applicable]. No nipples, areolas, blushing, genitals, wounds.

All (<id>)'s illustrAppearance should only be copied from the tracked items [Main Tracker] and/or [Main Cast]. IF  the (<id>) does not have an entry in "Main Tracker' or "Main Cast", then DO NOT CHANGE ANY OF THEIR illustrAppearance information. ANY CHARACTERS' illustrAppearance descriptions CAN be updated ONLY if the outcomeDescription EXPLICITLY states that the (<id>) has undergone a permanent change to an appearance field listed above.
Format: IMAGE STYLE OVERRIDE | IMAGE GUIDANCE DATA | IWUpscaleFaceSmooth | IWBeautiful| | [NUMBER] PEOPLE/Person SHOWN [VERB] | (<id_1>): [Data] | (<id_2>): [Data]

DO NOT OUTPUT BRACKETS [] OR SLASHES /. 

**illustrClothes:**
Restrictions: If naked → "Completely Naked-" + visible intimate details. If partially exposed → outfit AND detailed exposed part descriptions.
Format: (<id>): [Details] | (<id>): [Details]

**illustrExpressionPosition:**
Restrictions: Subjects, NPCs, or their body parts are strictly forbidden from Director Info. This field handles poses, positioning, ALL temporary appearance descriptors.
Format: (<id>): [Pose/Action/Emotion/Temporary effects/wounds/interaction/wet/magic/combat] | (<id>): [Details]

**illustrClothesChanged:**
If illustrSubject changes clothes in outcomeDescription → true. Otherwise false.
Format: (<id>): [Data] | (<id>): [Data]

**illustrAppearanceChanges:**
If PERMANENT change to eye/hair color/style, body type, bust, skin, non-human feature → summarize.
Format: (<id>): [Data] | (<id>): [Data]

**illustrSetting:**
Restrictions: THREE layers required. Named characters forbidden.
Format:
-- Background: [distant environment, materials, textures, lighting, shadows, time of day]
-- Midground: [immediate environment, materials, textures, lighting, shadows, effects]
-- Foreground: [minimal partial obstructions if present—bedsheets, furniture edges, etc. Keep extremely minimal. Write "None" only if absolutely nothing obstructs.]

</illustrInstructions>