KNOWN STATE — INDEPENDENT INFORMATION COLLECTION

We use KNOWN STATE because this site is an independent collection of information from cited sources and recorded observations as of the stated review date. This collection is in no way official and is not documentation, guidance, or endorsement issued by OD or its operators. Information and account behavior can change.

INDEPENDENT WEBSITE — NO AFFILIATION

duhployed.com is independently owned and operated. It is not affiliated with, associated with, owned by, operated by, sponsored by, endorsed by, or approved by OD, Dream Studio USA, Inc., or any of their parent companies, holding companies, subsidiaries, affiliates, or related entities.

Third-party trademarks belong to their respective owners. Read the independence and trademark notice.

OD / learning library
Download document ↓Find related resources →

Fluidity generator Quick Start

Inputs, manual placement, and illustrative examples for the matching generator.

Prompt structure reviewed. Efficacy unverified; limited native observations available. Read the native trial report.

The manifest review_status is a retained source-export label; it is separate from current native trial coverage.

Native efficacy of this exact published file is unverified. The report distinguishes manually authored image diagnostics, historical generator replies and exact-version instruction trials.

Role
Application guide
Category
CPD generators
Revision
1.2
Target / route
Read alongside the generator
Copy destination
Read the guide. Generator instructions stay in the authoring LLM; only the completed character output goes into CPD.
Primary pack
CPD authoring generators

Before copying

Read the stated destination and application notes first. They identify whether to copy complete instructions into an authoring LLM, paste its generated output, or read a guide. Replace requested placeholders. Nothing in the template runs automatically or performs an OD generation.

  • Examples demonstrate document structure; they are not observed native image results.
  • Text structure and application reviewed; no live OD generation or preservation benchmark was run.

Published text

OD FLUIDITY CPD GENERATOR - QUICK START
Version 1.2 | Complete-artifact delivery revision October 31, 2026

DEDICATED AUTHORING AND RAW OUTPUT CHECK
Use a separate authoring LLM or a dedicated OD authoring chat. Back up that chat's existing Custom Instructions, then replace conflicting inherited roleplay instructions with one complete generator. Keep the authoring task separate from story memories. A field placement or mode label does not guarantee instruction priority or output compliance.
Repeat the chosen scope and delivery format in each compilation request, including follow-ups. Use this request header outside the character payload:
Mode: CANONICAL CHARACTER. Return the complete artifact in this reply, even if unchanged. Do not refer me to an earlier result.
Format: exactly three nonempty comma-separated blocks separated by exactly two standalone uppercase BREAK lines, with a blank line before and after each separator; no headings, quotes, dialogue, thoughts or explanation.
Target: Mira, fictional adult woman aged 30, warm brown skin, brown eyes, oval face, lean build, black shoulder-length braids.
For a temporary profile, change Mode to CURRENT-FORM and repeat the complete identity plus the accepted change and coverage. Do not paste the request header, generator instructions or a status reply into CPD.
Inspect native Copy text before using a result. The raw result must contain exactly two standalone BREAK lines and three nonempty blocks. Reject missing artifacts and added prose rather than treating a claim of completion as completion. Display styling and raw copied text can differ.
LIMITED NATIVE OBSERVATIONS, OCTOBER 31, 2026
In one Eclipse Genius session displaying FREE, Fluidity v1.1 in Roleplay mode added prose and omitted BREAK lines. Changing only to Texting yielded correctly separated raw canonical and current-form outputs. Texting was marked SUNSETTING; its future availability and behavior are not guaranteed. Realistic anime v1.1 also returned references to earlier results on repeated requests. An added output guard helped one canonical retry but a later current-form request still returned a status message. These limited text observations motivated the complete-artifact rule and explicit request header; they do not establish a universal fix or native image fidelity. Recheck the actual copied artifact in the current model, mode and setup.

CHOOSE THE OUTPUT SCOPE FIRST
Request canonical CHARACTER for permanent identity, CURRENT-FORM for one accepted temporary appearance, or STYLE-ONLY for a separate rendering preset. Only a deliberately canonical CHARACTER result replaces CPD in the placement steps below. Keep CURRENT-FORM output with its actual scene handoff and reconcile saved appearance inputs manually. If mixed form data leaves the destination unclear, resolve that choice before compiling.

1. Paste the complete OD_Fluidity_CPD_Prompt_Generator.txt into a dedicated LLM chat or its instruction field.
2. Send one or two paragraphs describing the character and desired consistent rendering. State whether style-engine instructions are already active. Identify the features that matter most.
3. For a complete CHARACTER result, replace the intended CPD on an Anime: Fluidity character, keeping native appearance presets consistent. Save exact supplied size labels/measurements in the character's canonical notes or appropriate fields before replacing their only record. Generator instructions do not belong in CPD.

A STYLE-ONLY result is a rendering preset, not a whole-character replacement. To apply it to an existing card, supply the current CPD and request a complete merged replacement. This preserves identity/body/hair and restores exactly two separators. Do not paste three style blocks over appearance data or blindly append extra blocks.

The output has exactly two standalone BREAK separators. Most facts remain unweighted; selected related priorities receive modest numerical emphasis. The convention targets clearer instructions and fewer conflicts, not guaranteed output stability. A permanent CPD light treatment affects future images, so fixed time/source/location is omitted by default.

EXAMPLE INPUT - CHARACTER
An existing realistic-anime engine handles rendering. She is a fictional adult woman with deep brown skin, amber eyes, an oval face, a lean athletic upper body, narrow waist, and broad hips. Her natural black hair is worn in two long braids with copper cuffs. The braids and skin texture are the details I most want to retain. Tonight she is standing in rain wearing a red coat; that is only this scene.

ILLUSTRATIVE CANONICAL OUTPUT
(material-specific skin texture, selective focal detail:1.1), coherent light and shadow

BREAK

adult woman, deep brown skin, amber eyes, oval face, lean athletic upper body, narrow waist, broad hips

BREAK

black hair, (two long braids:1.1), copper hair cuffs

The one-scene rain, coat, and pose do not become permanent appearance. The rendering engine supplies the main style; the first block specifies compatible detail priorities.

STYLE-ONLY INPUT ALSO WORKS
Explicitly request a separate rendering preset if that is all you need. The output groups style/linework, light/color, and material/focal detail without inventing physique or hair. For direct entry on an existing character, provide its current CPD for a merged CHARACTER result instead.

TUNING
Inspect the actual generated prompt through Remix or More -> Details when available. If a trait is ignored, check its spelling and competing presets before increasing emphasis. Compare several outputs while changing one instruction at a time. Anatomy drift should prompt a conflict/reference review, not automatic breast, waist, or hip weights. Image editing requires the actual source and a separate preservation-focused edit instruction.

SOURCE BASIS
RansomDekafenya's supplied Fluidity Anime guide and its append-only research supplement informed the layout. The generator uses the updated distinction between documented syntax and community observations; it does not enforce an unverified 12-15-weight ceiling, double-weight accounting, or guaranteed BREAK isolation.
Known State reference: https://guide.ourdream.ai/characters/image-prompting
Permanent appearance and scenes: https://guide.ourdream.ai/characters/photo-shoot-and-gallery

The example is illustrative, not an OD generation test. Use the generator file as the operative instructions.

ACCEPTED FORM, COVERAGE AND SOURCE RECORD
Choose canonical permanent identity or an explicitly requested temporary current-form profile before generating text. For a changing character, supply the accepted transition, form/stage, complete relevant visible appearance and unchanged anchors. A partial mixture is intentional only when defined. Do not blend every form, infer new anatomy or copy obsolete skin/hair back over a changed feature.
Only a canonical CHARACTER result is intended to replace permanent CPD. A CURRENT-FORM profile stays with the actual scene/image handoff unless a permanent revision is explicitly intended. A saved character or preset may still inject an older appearance: inspect and reconcile those inputs rather than assuming an ED block, snapshot or tracker overrides them.
The skin-tone automatic-clothing-coverage workaround was withdrawn by its supplied author after confounded initial tests and contrary clean-character results. Keep stable identity separate from what the current garments and crop reveal; field placement does not guarantee occlusion. An affection score, stage label or SCE heading does not establish permission, undressing or a completed action.
For editing, select actual accepted source pixels. Record source -> crop -> attempted edit -> accepted result, including failed passes and each displayed charge, outside the generation prompt. State the change and its explicit preservation exception; inspect each subject, pose and background geometry separately. Multiple-image identity/composition transfer is only an experiment when the route actually accepts those inputs.
The supplied abbreviation PD is not expanded; do not guess its field. Replacing smoke or mist with vapors is an untested lexical suggestion, not a verified fix. Emphatic copy wording, nested weights and a selected success do not isolate a reliable cause. A still crop is not camera motion, and native video output must be distinguished from later external splicing or audio work.
Current-form worksheet: https://duhployed.com/learn/modules/current-form-continuity.html
Tracker design review: https://duhployed.com/learn/modules/affection-tracker-review.html
Scenario worksheet: https://duhployed.com/learn/modules/scenario-engineering.html

CURRENT-FORM INPUT EXERCISE
Canonical Mira is a fictional adult archivist with a warm brown face, brown eyes, black shoulder-length braids and a lean build. The accepted change affects only her hands: they are gray granite. Her face, hair and proportions remain unchanged; her navy jacket and cream shirt stay on. Request CURRENT-FORM, not a permanent CPD rewrite. Inspect that the resolved appearance profile includes the granite hands without turning her face or hair into stone, and that the canonical record remains separate. Use the matching generator's actual output contract: the Body Horror v2.2 default is one unweighted payload, while Fluidity and realistic anime retain their explicitly documented three-block convention. This is an authored fixture, not an observed generation.

9502 bytes · SHA-256: 8ade79d87b5e1aa869c7aa95ed69a6134cee15880da09d4b5fa831e200db120d