Home
Design
Art
About
Scaling Product Photography with Weave: An AI Character-to-Campaign Pipeline
What is this
An explorational project showing how to use generative tools like Weave to create workflows that scale with consistency. Using LLMs and system prompts to automate the asset creation and reducing slop.
Swapping SKU options without losing details and being consistent. Needing only minimal retouching.
Examples of final outcomes created using the workflow. The product and talent are largely consistent across the outputs. Product used - Gentle Monster (For illustration purposes only).
How does it scale
Character and styling are separated into swappable nodes, so one identity can be re-styled or re-shot without rebuilding the pipeline.
Having a system and working in steps helps the team to easily adapt this workflow without breaking patterns.
Generating a talent with consistent AI models as the foundation for every downstream asset. I’ve set custom parameters describing the talent I want to generate
Layered styling nodes - outerwear, base, bottom, shoes; each independently swappable. The nodes are customisable and does not require any prompting.
Styling generator walkthrough built using the selected talent as a base
Character sheet output - Locking identity, styling and talent before scaling to production
A system prompt defines the style rules. It feeds into an LLM node that generates image prompts. Character sheets and product references go to the image models for consistency. I just type my idea, the rest is automated
All the inputs are attached throughout to generate consistent outputs.
Generated image input into Kling 3 to add motion to the static output
Attached character sheet and product to Kling as Elements to keep outputs consistent. These references give the model all angles it needs, so it only uses what's provided rather than generating new details
Final campaign-ready outputs with talent, styling, and shots generated end-to-end. Consistent output with talent + product being used to produce assets at scale
Future potential
Multi-SKU campaign batches auto-generated per drop, localising content at scale across video, social, UI and print.
A systems-based workflow makes production scalable and repeatable. Applicable to illustrations, on-brand assets, icons and more.
Link to workflow: https://app.weavy.ai/flow/TruUYLF9HsIXqmkY8aj6MC
Other Projects
See Project
→
See Project
→
See Project
→
Let’s work together
CONTACT
girivarshan.work@gmail.com
+61 468 469 067
SOCIAL
© 2026 Girivarshan Balasubramanian. All rights reserved.
Home
Design
Art
About
Scaling Product Photography with Weave: An AI Character-to-Campaign Pipeline
What is this
An explorational project showing how to use generative tools like Weave to create workflows that scale with consistency. Using LLMs and system prompts to automate the asset creation and reducing slop.
Swapping SKU options without losing details and being consistent. Needing only minimal retouching.
Examples of final outcomes created using the workflow. The product and talent are largely consistent across the outputs. Product used - Gentle Monster (For illustration purposes only).
How does it scale
Character and styling are separated into swappable nodes, so one identity can be re-styled or re-shot without rebuilding the pipeline.
Having a system and working in steps helps the team to easily adapt this workflow without breaking patterns.
Generating a talent with consistent AI models as the foundation for every downstream asset. I’ve set custom parameters describing the talent I want to generate
Layered styling nodes - outerwear, base, bottom, shoes; each independently swappable. The nodes are customisable and does not require any prompting.
Styling generator walkthrough built using the selected talent as a base
Character sheet output - Locking identity, styling and talent before scaling to production
A system prompt defines the style rules. It feeds into an LLM node that generates image prompts. Character sheets and product references go to the image models for consistency. I just type my idea, the rest is automated
All the inputs are attached throughout to generate consistent outputs.
Generated image input into Kling 3 to add motion to the static output
Attached character sheet and product to Kling as Elements to keep outputs consistent. These references give the model all angles it needs, so it only uses what's provided rather than generating new details
Final campaign-ready outputs with talent, styling, and shots generated end-to-end. Consistent output with talent + product being used to produce assets at scale
Future potential
Multi-SKU campaign batches auto-generated per drop, localising content at scale across video, social, UI and print.
A systems-based workflow makes production scalable and repeatable. Applicable to illustrations, on-brand assets, icons and more.
Link to workflow: https://app.weavy.ai/flow/TruUYLF9HsIXqmkY8aj6MC
Other Projects
See Project
→
See Project
→
See Project
→
Let’s work together
CONTACT
girivarshan.work@gmail.com
+61 468 469 067
SOCIAL
© 2026 Girivarshan Balasubramanian. All rights reserved.
Home
Design
Art
About
Scaling Product Photography with Weave: An AI Character-to-Campaign Pipeline
What is this
An explorational project showing how to use generative tools like Weave to create workflows that scale with consistency. Using LLMs and system prompts to automate the asset creation and reducing slop.
Swapping SKU options without losing details and being consistent. Needing only minimal retouching.
Examples of final outcomes created using the workflow. The product and talent are largely consistent across the outputs. Product used - Gentle Monster (For illustration purposes only).
How does it scale
Character and styling are separated into swappable nodes, so one identity can be re-styled or re-shot without rebuilding the pipeline.
Having a system and working in steps helps the team to easily adapt this workflow without breaking patterns.
Layered styling nodes - outerwear, base, bottom, shoes; each independently swappable. The nodes are customisable and does not require any prompting.
Generating a talent with consistent AI models as the foundation for every downstream asset. I’ve set custom parameters describing the talent I want to generate
Styling generator walkthrough built using the selected talent as a base
Character sheet output - Locking identity, styling and talent before scaling to production
A system prompt defines the style rules. It feeds into an LLM node that generates image prompts. Character sheets and product references go to the image models for consistency. I just type my idea, the rest is automated
All the inputs are attached throughout to generate consistent outputs.
Generated image input into Kling 3 to add motion to the static output
Attached character sheet and product to Kling as Elements to keep outputs consistent. These references give the model all angles it needs, so it only uses what's provided rather than generating new details
Final campaign-ready outputs with talent, styling, and shots generated end-to-end. Consistent output with talent + product being used to produce assets at scale
Future potential
Multi-SKU campaign batches auto-generated per drop, localising content at scale across video, social, UI and print.
A systems-based workflow makes production scalable and repeatable. Applicable to illustrations, on-brand assets, icons and more.
Link to workflow: https://app.weavy.ai/flow/TruUYLF9HsIXqmkY8aj6MC
Other Projects
See Project
→
See Project
→
See Project
→
Let’s work together
CONTACT
girivarshan.work@gmail.com
+61 468 469 067
SOCIAL
© 2026 Girivarshan Balasubramanian. All rights reserved.