Models & Presets
Fluxara uses preset-based execution to keep media workflows stable, repeatable, and easier to compare across runs. Presets act as the main configuration layer between user input and model execution.
Why presets exist
In many internal workflows, consistency matters more than unlimited flexibility. A preset captures a known-good configuration so that teams can rerun the same workflow without rebuilding execution parameters every time.
This makes it easier to compare outputs, reduce configuration drift, and maintain a predictable relationship between input, workflow, and final artifact.
Preset structure
A preset may define a combination of model, inference, and output settings. The exact structure may evolve during the beta period, but common fields include:
Model identifier
Selects the model family or specific model variant used during generation or transformation.
Resolution defaults
Defines the expected output dimensions or baseline size profile for a workflow.
Inference steps
Sets the number of execution steps used during generation to balance speed and output quality.
Sampler configuration
Defines the sampling method or execution profile used by the selected model.
Prompt templates
Supplies reusable positive or negative prompt scaffolding for a given workflow type.
Export profile
Defines how the final artifact is written, including format and related output metadata behavior.
Model role in the workflow
The selected model determines the main output behavior, but it does not operate in isolation. In Fluxara, the model works together with:
- the selected preset
- the pipeline structure
- the user input or source asset
- execution options provided at job time
This layered structure helps separate stable defaults from per-job variation.
Preset categories
Generation presets
Used for prompt-driven output creation where the main input is text and the output is a new visual artifact.
Transformation presets
Used when an existing image or asset is passed through a structured modification workflow.
Review presets
Used to generate more comparable outputs across repeated runs for internal evaluation and selection.
Export presets
Used to standardize artifact formatting, size, and final delivery expectations across workflows.
Reproducibility goals
Presets are a key part of making Fluxara reproducible. They allow teams to hold configuration stable while changing only one variable at a time, such as:
- prompt wording
- source image
- pipeline stage selection
- output resolution override
This is especially useful in internal review workflows where outputs need to be compared across multiple runs or evaluated by more than one person.
Example preset logic
In practice, a preset combines these elements into a reusable profile. A job can then reference the preset directly instead of redefining all parameters for every run.
Current beta constraints
- Preset format may evolve during beta.
- Not all configuration options are exposed to end users.
- Preset editing may remain limited in early onboarding phases.
- Model availability may differ across approved beta environments.
Design principles
Stability first
Presets should reduce surprises and help workflows behave consistently across repeated runs.
Controlled flexibility
Teams can vary inputs and workflow choices without rebuilding the full execution stack each time.
Comparability
Outputs should be easier to evaluate side by side when presets are held constant.
Operational reuse
Good defaults should be reusable across projects, batches, and internal review cycles.