Skip to content

getting-started.md

Latest commit

 

History

History
239 lines (162 loc) · 8.48 KB

getting-started.md

File metadata and controls

239 lines (162 loc) · 8.48 KB
Raw
layout default
title Getting Started
nav_order 2

Getting Started

This guide walks you through setting up Batch Generate Image with any AI and running your first batch.

1. Install the app

Install Batch Generate Image with any AI from the Microsoft Store, then launch it from the Start menu.

The first launch opens a Getting Started dialog automatically. You can reopen it any time with the ? button in the header.

2. Set up your AI connection

Before generating anything, you need an API key from your chosen provider.

  1. Click the key icon in the top-right header
  2. Select a provider from the dropdown
  3. Click Get API key if you need to open the provider page
  4. Paste your API key
  5. Click Test connection
  6. Click Done

Your API key is encrypted and stored on your device using Windows DPAPI. It is sent directly to your provider.

Multiple connections: You can save up to 10 named connections. This is useful when you want one setup for OpenAI, another for OpenRouter, another for a cheaper FLUX batch model, and another for a custom endpoint. Use Set as active connection to switch.

Practical tip: OpenAI is the simplest first-run option. OpenRouter is a good second choice if you want one key for multiple model families. Fireworks AI and Together AI are useful for cheaper open-source image models, but some analysis features may be unavailable depending on the model.

3. Configure your persona (Step 1)

The persona is your reusable visual brief. It shapes the overall style of the batch before the row-specific description is added.

The app includes three starter personas:

  • Vivid Vera — bold colours, stronger contrast, dramatic energy
  • Minimalist Max — cleaner layouts, restraint, negative space
  • Fantasy Flora — painterly, atmospheric, richer storytelling mood

You can also create your own personas and edit:

  • artist type
  • visual mediums
  • lighting styles
  • color palettes
  • moods
  • composition rules
  • do / avoid guidance
  • quality requirements
  • 10 visual datapoints

To edit a persona, click Edit on the persona card.

Analyze sample images for persona suggestions

Inside the persona editor, use Analyse samples to pick a folder of your own images. The app will inspect that folder and suggest persona values that better match the style of those images.

This is separate from Step 2 style-reference analysis:

  • Sample analysis helps you improve the persona itself
  • Style-reference analysis guides a specific generation run

Global style controls in Step 1

Content Style — choose from 24 styles including Photorealistic, Cinematic, Watercolor, Storybook, Editorial Illustration, Coloring Book Line Art, Invitation / Stationery, and more.

Black & White — convert all outputs to greyscale.

Include Captions — when on, any caption field in your image list is rendered as visible text in the image. See Captions for layout and preview options.

4. Load style reference images (Step 2, optional)

Style-reference images help the app understand the look you want from a specific folder of inspiration images.

  1. Choose a folder containing .png, .jpg, .jpeg, .webp, or .gif files
  2. Click Analyze

Important behavior:

  • The app does not auto-analyze the folder
  • Until you click Analyze, prompt preview and generation ignore the reference folder
  • Your prompt, persona, template, and content style stay primary
  • The reference folder acts as soft secondary guidance only

After analysis, the app can show:

  • a short summary of the folder
  • whether the folder looks consistent or mixed
  • cue matches and gaps compared with the active persona
  • a 10-point visual comparison between reference traits and persona traits

Provider note: Style-reference analysis needs a vision-capable chat model. Image-only generation models such as FLUX can still generate final images, but they cannot inspect a folder by themselves. Some providers can temporarily fall back to a separate vision model for analysis; others cannot.

Re-run when needed: If you change the folder, change the active provider, or change the model, click Analyze again. Balanced and Sample generation focus require a current analysis.

5. Prepare your image list (Step 3)

The image list is a CSV or YAML file with one row per image you want to generate. Each row becomes one image request.

Generate a list with AI

Click Generate CSV in Step 3 to open the built-in list generator. Enter the category, subject, style, mood, and notes, then generate a ready-to-use CSV.

The generated file is saved to:

Documents\BatchGenImage\generated-lists\

It is also loaded into the app automatically.

You can optionally turn on Add AI captions column in the CSV dialog. This adds a caption field for each row so caption overlays are ready to use.

Model note: AI CSV writing uses a chat-capable model. If your active connection is an image-only model like FLUX, switch to a chat-capable connection for AI-written CSV rows and captions. If not, the app can still save a simpler local starter CSV.

Try an example first

Click Load Example in Step 3 to load an example file for the active template.

Field reference

Field Aliases accepted Required
title name, topic Yes
description desc, prompt, content Yes
subject main_subject, focal, focus No
mood tone, feeling, atmosphere, vibe No
tags keywords, styles, style, keyword No
notes note, guidance, context, instructions No
caption text, overlay, label, tagline No

CSV format

Use semicolons inside the tags field if you want multiple tags:

title,description,subject,mood,tags,notes,caption
"Summer Festival Poster","Vibrant outdoor music festival at sunset with crowd and stage lights","Concert stage","Energetic and festive","music;festival;summer","Warm golden hour lighting","Summer Beats 2025"
"Tech Conference","Sleek modern conference poster with geometric shapes and neon accents","Abstract tech shapes","Professional and futuristic","technology;minimal;neon",,

Quoted headers are supported too:

"title","description","caption"
"Summer Festival Poster","Vibrant outdoor music festival at sunset","Summer Beats 2025"

YAML format

For YAML, use arrays for tags:

- title: "Summer Festival Poster"
  description: "Vibrant outdoor music festival at sunset with crowd and stage lights"
  subject: "Concert stage"
  mood: "Energetic and festive"
  tags:
    - "music"
    - "festival"
    - "summer"
  notes: "Warm golden hour lighting"
  caption: "Summer Beats 2025"

- title: "Tech Conference"
  description: "Sleek modern conference poster with geometric shapes and neon accents"
  subject: "Abstract tech shapes"
  mood: "Professional and futuristic"
  tags:
    - "technology"
    - "minimal"
    - "neon"

Only title and description are required.

6. Run your first generation

  1. Choose your template
  2. Confirm your output folder
  3. If you analyzed a reference folder, choose the Generation Focus
  4. Click Preview Prompt
  5. Click Generate

Large batches show a cost confirmation before the app starts.

Templates

The app includes 12 built-in templates plus a custom template. Current built-in options include:

  • Poster
  • Social Media Post
  • Banner / Cover
  • Book Cover
  • Book Illustration Page
  • Children's Book Illustration
  • Storybook Spread
  • Coloring Book Page
  • Invitation / Save the Date
  • Product Mockup
  • Album Art

Generation Focus

When a current Step 2 analysis exists, you can choose:

Option Effect
Persona Follow the persona only. Reference guidance is ignored.
Balanced Share influence between persona and reference folder.
Sample Lean more strongly toward the reference folder while still respecting persona rules.

If no current analysis exists, only Persona is available.

Next steps

  • Try different personas on the same image list to see how artistic direction changes results
  • Use Analyse samples inside the persona editor if you want the persona to better match your existing image library
  • Use Prompt Preview before large batches
  • Explore the Templates editor in the header to customize dimensions, quality, and prompt guidance
  • Add a caption column or use Add AI captions column in the CSV generator if you want visible text on the final images