AI Modify

AI Modify is a generative AI feature that transforms photos taken at the booth. It works on the original photo — the people, faces, poses, and group layout stay intact — and AI applies changes on top of that foundation, guided by your prompt. This makes results faster, more predictable, and far better suited for live events than fully generative alternatives.

---

Two Ways to Use AI Modify

1. Specific "Photoshop-Style" Edits

Small, controlled changes where the AI treats the photo as a base layer: background swaps, lighting adjustments, outfit changes, or studio-quality corrections. These are fast, low-risk, and work well even with larger groups.

2. Creative AI Transformations

Bigger changes where the AI builds an entirely new scene around the people — from brand activations to fantasy environments. AI Modify gives you control across this entire range, from subtle edits to bold creative scenes.

The newer models handle graphic-design output — trading cards, magazine covers, collectible packaging — including legible text inside the image.

AI Models

These are the models you can select in the Workflow Builder. The right choice depends on your group size, prompt complexity, event pace, and budget.

Model	Basic	Pro	Avg Speed	Resemblance	Best For
Grok Imagine	$0.10	$0.05	~10–15 sec	Good	Fastest and lowest-cost model; bold, expressive results ideal for high-volume events
Nano Banana 2 Lite	$0.15	$0.10	Under 10 sec	Good	The speed option. Great when throughput matters most and the photo is one person
Nano Banana	$0.16	$0.08	~12 sec	Good	Fast, small edits: background swaps, effects, props, studio looks
GPT-Image 1.5	$0.16	$0.08	~30 sec	Strong	Strict instruction following: when the layout, pose, and composition rules must be respected
GPT-Image 2	$0.16	$0.08	~20-80 sec	Strongest	Highest quality and instruction precision available — cleanest composition, fewest retries
Nano Banana 2	$0.28	$0.20	~10–25 sec	Strong	Creative transformations where visual quality and likeness both matter

Pricing shown as cost per generation, Basic and Pro plan. See current rates inside Foto Master Cloud.

Group Photos

Standard AI tools work well for one or two people. Groups are where most systems struggle — common issues include generic-looking faces, loss of individual features, and uneven results across the group.

Face limits are real, and they scale with model tier. Lower-tier and faster models hold resemblance well for a single person but drift as the group grows. Higher-tier models — GPT-Image 2 in particular — hold likeness noticeably better across more faces.

Choosing the Right Model

• Speed and volume — Grok Imagine or Nano Banana 2 Lite. Fastest generation, lowest cost, ideal for full-day events and long lines.
• Best all-round default — Nano Banana 2. Fast enough for live events with high quality and strong resemblance.
• Strict prompt adherence — GPT-Image 1.5 or GPT-Image 2. Best when the layout, composition, and specific scene details must match exactly.
• Maximum quality — GPT-Image 2. The highest output quality and strongest resemblance in the lineup, for brand activations, VIP experiences, and client presentations.
• Text inside the image — Grok Imagine or GPT-Image 2. Both handle signage, captions, and graphic-design layouts well.

Step Settings

AI Modify is added as a Cloud step inside the Workflow Builder. Click the gear icon on the node to open its settings.

Model

Select the model from the dropdown. You can change it at any time before syncing — model choice is not locked to the workflow.

Prompt

Describe what you want to change. Keep it concise: the environment, style, what stays the same, and any specific elements. See How to Write Better Prompts for AI Modify and AI Combine for detailed guidance.

The + button inside the Prompt field inserts dynamic values instead of fixed text:

• Survey responses — the guest types their own description at the booth and the AI uses it. See Setting Up User-Generated AI Prompts in FMX via Survey Feature.
• Selection screen values — the guest picks from options you define, and each choice feeds a different prompt.

Tip: when using a dynamic prompt, lead with a short fixed phrase to steer the AI — for example, "Dress the people in this photo as…" — then append the guest's input.

AI Error Handling

Turn on Go to specific step on AI Error to route the workflow somewhere else when a generation fails, instead of leaving the guest at a dead end. Point it at a retry, a message screen, or a plain photo path. Worth configuring on any unattended booth.

Description (optional)

A short label describing what the node does. It appears as a collapsible label under the node in the graph. Useful when a workflow has several AI steps or when someone else will maintain it.

Testing

To test prompts without running the booth, use AI Playground in Foto Master Cloud. Compare models on the same photo and prompt before you commit to one for the event.

Getting Started

Add the AI Modify step to your workflow, choose the model, and write your prompt during workflow design. Save the workflow and sync it to the booth. Guests trigger it at the booth and receive the result in seconds.