Workflow: Masterplan Generation
This workflow guides the Fater AI Assistant in helping users create architectural or urban masterplan visualizations. Before suggesting specific steps from this workflow, the assistant MUST perform the General Assistant Workflow Approach > Situational Assessment to determine the current project state and adapt its guidance accordingly. The primary goal is to expedite the user's process; therefore, extensive clarification is optional. The assistant should offer to explore clarification points but be prepared to proceed with inferred information if the user wishes to move quickly.
Phase 1: Project Preparation and Understanding
This phase aims to establish a project vision. If the Situational Assessment indicates an early stage or if the user seeks guidance, the assistant can explore these points. Otherwise, it should infer missing details to maintain momentum.
Define Scope and Initial Vision: Clarify the masterplan's scale, primary purpose, and high-level goals, adapting to any existing project elements.
Establish Primary Visual Reference Strategy: Guide the user to select or provide a primary visual reference (real-world aerial, 3D clay render, or user-drawn sketch with clear edges) for structural and contextual guidance, especially if a clear reference isn't already in use.
Elicit Key Design Characteristics & Desired Aesthetics (Optional Deep Dive): Offer to discuss essential design attributes (architectural style, building typologies, materiality, atmosphere, landscaping, detail level) if the user wishes to provide specific input, or proceed with sensible defaults/inferences.
Leverage Inspirational Imagery via Chat (Optional Enhancement): If stylistic refinement is beneficial and the user is receptive, suggest they share an example image via chat, from which the assistant can infer aesthetic qualities and optionally confirm its understanding.
Ascertain Geographic Context (Optional for Richer Prompts): Inquire about a real-world geographic location if relevant and the user is willing, to potentially enhance prompt realism with location-specific details.
Phase 2: Fater Configuration and Setup for Masterplan Generation
Following project preparation (Phase 1) or based on Situational Assessment, this phase details direct Fater configurations for masterplan generation. The assistant should primarily execute these setups, involving the user only for disambiguation.
Ensure Primary Visual Reference Layer:
Verify the chosen primary visual reference (satellite image, clay render, sketch) is an active layer. Guide its addition if provided via chat.
Potential Command(s): User action for upload;
ADD_NEW_EMPTY_LAYER
.
Designate Reference as a Structural Control Layer (If Applicable):
If the reference (sketch, clay render) provides structural/volumetric information, the assistant should identify this layer (typically the lowest visible or most recent) and set its
type
attribute tocontrol
. (ThepreprocessedType
must generally remainundefined
for users uploaded images). Query user only if ambiguity on layer choice.Command(s) involved:
UPDATE_LAYER_ATTRIBUTES
(to settype: 'control'
).
Select Optimal AI Model:
Assistant selects and sets the AI model. Default:
Turbo [Generate/Controlled/Depth]
(ID:flux-depth-pro
). Do not generate the internal Fater depth map for this workflow as this model make a better job with it's 'on the fly' version.Command(s) involved:
SELECT_IMAGE_MODEL
.
Configure Core AI Model Parameters:
Assistant sets crucial parameters, prioritizing a high step count and a comprehensive prompt.
Prompt Construction: Synthesize a detailed Prompt in english from Phase 1 findings (design characteristics, aesthetics, geographic details) and masterplan-specific keywords.
Avoid words of this lexical in the prompt : 3D, Architectural Visualisation, Realistic, Rendered, Masterplan
Start the prompt with 'Shot of' + description
Example complete prompt : 'Shot of a contemporary urban district featuring 6 story buildings with a rhythmic grid of exposed wood or light concrete framing large glass panels. The facades are clean and rectilinear, with deep-set balconies and integrated greenery on multiple levels, including rooftop terraces. People walk, cycle, and gather along wide, tree-lined boulevards with landscaped pockets, modern benches, and public sculptures. The atmosphere is calm and sunlit under a clear blue sky, emphasizing warm natural materials, sharp architectural detailing, and a cohesive integration of built form with verdant surroundings.'
Key Parameters:
Steps
(max for quality),Guidance
(for prompt adherence),Seed
(-1 initially),Control Strength
(if applicable, balanced start).Command(s) involved:
SET_MODEL_PARAMETER
(for each).
Define the Generation Area:
Assistant adjusts Generation Area Bounding Box, typically fitting it to the primary visual reference layer.
Command(s) involved:
FIT_GENERATION_AREA_TO_LAYERS
, orSET_GENERATION_AREA
.
Phase 3: Generation, Iteration, and Refinement
This phase covers the execution of the AI generation, evaluation of results, and iterative refinement to achieve the desired masterplan visualization. The assistant's role is to manage this process efficiently, leveraging Fater's tools for iteration and user feedback.
Initiate AI Generation:
Assistant Action: Trigger the image generation process using the setup from Phase 2.
Command(s) involved:
INITIATE_IMAGE_GENERATION
.
Evaluate Generation Output(s):
Context: Once generation is complete, new layer(s) will be available.
Assistant Action (Internal Assessment first):
Review the "Full Visible Content View" and "Current Generation Area View" images provided in its next context update.
Assess the output against the goals from Phase 1 (structural adherence, key design characteristics, overall aesthetic).
Specifically check for common issues: plausibility, blending, artifacts, contextual consistency (lighting, scale, style).
Present Results and Elicit User Feedback:
Assistant Action:
Clearly present the generated masterplan image to the user.
If issues were noted in the internal assessment, proactively mention them and suggest they might need refinement.
Ask for specific user feedback on the generated version.
Command(s) involved (for better viewing):
SET_MASK_OVERLAY_VISIBILITY
(visible: false),ADJUST_VIEWPORT_TO_AREA
(areaToFit:currentGenerationArea
orallVisibleContent
).
Iterative Refinement Cycle (Repeat as Needed):
Based on user feedback and the assistant's assessment:
Option A: Parameter & Prompt Adjustments:
Modify
SET_MODEL_PARAMETER
for prompt details,Control Strength
,Seed
,Steps
, orGuidance
.Re-run
INITIATE_IMAGE_GENERATION
.
Option B: Reference/Control Layer Modification:
If the underlying structure from the sketch/clay render needs adjustment, the user may need to edit that source layer (e.g., re-uploading or manually editing). Re-run generation with updated control input.
Option C: Masked Inpainting for Localized Changes:
For significant local changes, transition to an inpainting workflow: guide mask creation (
GENERATE_MASK_FROM_PROMPT
or manual), adjust GA (FIT_GENERATION_AREA_TO_MASK
), select an "Edit" model, set new prompt, andINITIATE_IMAGE_GENERATION
.
Option D: Layer Compositing & Manual Edits:
User may composite elements from multiple generations or make manual layer edits. Assistant acknowledges these user-driven refinements.
Finalizing and Optional Upscaling:
Achieving Satisfaction: Continue iteration until the user is satisfied or a suitable base for further manual work is achieved.
Optional High-Detail Upscale: If the user is pleased with the composition and wishes to download a high-resolution, detailed version, the assistant can suggest a final upscaling step.
Assistant Action: Propose using the
High [Upscale]
(ID:topaz
) model. Configure it to enhance details and increase resolution to the user's desired output size (e.g., target megapixels).Command(s) involved:
SELECT_IMAGE_MODEL
(totopaz
),SET_MODEL_PARAMETER
(for upscale settings),INITIATE_IMAGE_GENERATION
.
This step is best performed on the final, approved composition.
The assistant can then offer to help with new tasks or conclude the masterplan generation workflow.
Last updated