03 / CUSTOM VISUAL MODELS
Custom LoRA training
Define a visual direction, prepare the data, train a model and compare the outputs. Turn scattered references into a repeatable creative capability.
Move from a promising image to a considered visual direction.
For design and content teams developing a specific style, brand image language or visual series.
Defined deliverables.
One visual-style LoRA model file
Evaluation images, prompt guidance and usage notes
SCOPE & BOUNDARIES
Clear boundaries. Better projects.
The base scope targets one visual style. You must have rights to use the training material. Base-model licensing, data quality and intended use are checked before training. GPU costs are separate; identical results across generations are not guaranteed.
Starting prices cover the base scope. A written quote follows discovery. Model/API usage, GPU and cloud costs, asset licenses, additional languages and ongoing maintenance are separate. Taxes and payment terms are specified in the quote.
Compare service pricingGood work has a clear path.
- 01
Understand
Define the goal, audience, available material and constraints.
- 02
Scope
Agree on deliverables, technical direction, acceptance criteria and a written quote.
- 03
Make
Turn the core idea into something tangible, then refine it through feedback.
- 04
Hand over
Review against the agreed scope and deliver the code, files and operating notes.