images2slides

Convert static infographic images into editable Google Slides. Provide your infographic images, and the system automatically analyzes them using Vision-Language Models (VLMs) to extract text and visual elements, then reconstructs them as native, editable slide elements.

Demo

Click the image above to watch the demo video

Features

• End-to-End Pipeline: Just provide images - VLM analysis, layout extraction, and slide creation are all automated
• Multi-Provider VLM Support: Google Gemini, OpenAI, Anthropic, and OpenRouter
• Multi-Slide Presentations: Convert multiple infographics into a single presentation
• Flexible Configuration: Configure via .env file, environment variables, or CLI arguments
• Coordinate Transformation: Map pixel coordinates to slide points with aspect-ratio-preserving fit
• Slide Reconstruction: Create editable text boxes and placed images via Google Slides API

Quick Start

# 1. Install
cd images2slides
uv sync

# 2. Configure (copy and edit .env.example)
cp .env.example .env
# Edit .env with your API keys and credentials

# 3. Convert images to slides
uv run images2slides convert \
  --image slide1.png \
  --image slide2.png \
  --title "My Presentation"

Note: All CLI commands must be run with uv run prefix (e.g., uv run images2slides ...), which executes them within the project's virtual environment.

---

Installation

Prerequisites

• Python 3.11+
• uv package manager

Install

cd images2slides
uv sync

---

Configuration

All configuration can be set via:

1. .env file (recommended)
2. Environment variables
3. CLI arguments (override .env and env vars)

Step 1: Copy the Example Configuration

cp .env.example .env

Step 2: Configure VLM Provider

Edit .env and set your VLM provider and API key:

# Choose your provider: google, openai, anthropic, or openrouter
VLM_PROVIDER=google

# Set the API key for your chosen provider
GOOGLE_API_KEY=your-api-key-here

VLM Provider Options

Provider	Default Model	API Key Variable	Get API Key
google (default)	gemini-3-pro-preview	GOOGLE_API_KEY	Google AI Studio
openai	gpt-5.2	OPENAI_API_KEY	OpenAI Platform
anthropic	claude-opus-4-5	ANTHROPIC_API_KEY	Anthropic Console
openrouter	qwen/qwen3-vl-235b-a22b-instruct	OPENROUTER_API_KEY	OpenRouter

To use a specific model, set VLM_MODEL in .env:

VLM_MODEL=gemini-3-pro-preview

Step 3: Configure Google Slides API Access

You need to authenticate with Google to create presentations. Choose one of these methods:

Option A: OAuth 2.0 Client Secret (Recommended for Personal Use)

This method opens a browser window for you to log in with your Google account. Best for local development and personal use.

How to get client_secret.json:

1. Go to Google Cloud Console
2. Create a new project or select an existing one
3. Enable the Google Slides API:
   • Go to "APIs & Services" > "Library"
   • Search for "Google Slides API"
   • Click "Enable"
4. Create OAuth credentials:
   • Go to "APIs & Services" > "Credentials"
   • Click "Create Credentials" > "OAuth client ID"
   • If prompted, configure the OAuth consent screen first (External is fine for personal use)
   • Select "Desktop app" as application type
   • Name it (e.g., "Slides Infographic")
   • Click "Create"
5. Download the credentials:
   • Click the download button next to your new OAuth client
   • Save the file as secrets/client_secret.json

Configure in .env:

CLIENT_SECRET_PATH=secrets/client_secret.json

On first run, a browser window will open for you to authorize the app. The token is cached for future use.

Option B: Service Account (Recommended for Automation/Servers)

This method uses a service account key file. Best for automated pipelines and server deployments. No browser interaction needed.

How to get service_account.json:

1. Go to Google Cloud Console
2. Create a new project or select an existing one
3. Enable the Google Slides API (same as above)
4. Create a service account:
   • Go to "IAM & Admin" > "Service Accounts"
   • Click "Create Service Account"
   • Name it (e.g., "slides-infographic")
   • Click "Create and Continue"
   • Skip the optional steps, click "Done"
5. Create a key:
   • Click on the service account you just created
   • Go to "Keys" tab
   • Click "Add Key" > "Create new key"
   • Select "JSON" format
   • Save the file as secrets/service_account.json

Configure in .env:

SERVICE_ACCOUNT_PATH=secrets/service_account.json

Important: Service accounts create presentations in their own Drive. To access them:

• Share the presentation with your email after creation, OR
• Set up domain-wide delegation to impersonate users

Which Should I Choose?

Use Case	Recommended Method
Local development	OAuth 2.0
Personal scripts	OAuth 2.0
CI/CD pipelines	Service Account
Server applications	Service Account
Shared team tool	Service Account with delegation

---

Usage

All commands are run with uv run to execute within the virtual environment.

Main Command: Convert Images to Slides

uv run images2slides convert \
  --image infographic1.png \
  --image infographic2.png \
  --title "My Presentation"

This command:

1. Analyzes each image with the configured VLM
2. Extracts text regions and image regions (icons, logos, charts)
3. Optionally uploads cropped image regions to GCS
4. Creates a new Google Slides presentation
5. Builds editable slides with text boxes and images

Including Image Regions

If your infographics contain icons, logos, or charts that the VLM detects as image regions, you need a Google Cloud Storage bucket to host them:

uv run images2slides convert \
  --image infographic.png \
  --title "My Presentation" \
  --gcs-bucket your-bucket-name

Or set GCS_BUCKET in your .env file. Without a GCS bucket, image regions will be skipped and you'll see a warning.

Output:

Step 1: Analyzing 1 image(s) with gemini-3-pro-preview...
  [1/1] Analyzing: infographic.png
         Found 5 text, 3 image regions

Step 2: Uploading 3 image region(s) to GCS...
  [1/1] Cropping 3 regions from infographic.png

Step 3: Connecting to Google Slides API...

Step 4: Creating presentation 'My Presentation'...

============================================================
SUCCESS!
============================================================
Presentation URL: https://docs.google.com/presentation/d/abc123/edit
Presentation ID:  abc123
Slides created:   1

CLI Options

uv run images2slides convert --help

Option	Description	Default
--image	Path to infographic image (repeatable)	Required
--title	Presentation title	"Infographic Presentation"
--page-size	Slide size: 16:9, 16:10, or 4:3	16:9
--provider	VLM provider	From VLM_PROVIDER or "google"
--model	VLM model	From VLM_MODEL or provider default
--gcs-bucket	GCS bucket for image regions	From GCS_BUCKET
--save-layouts	Save layout JSON files to directory	-
--client-secret	OAuth client secret path	From CLIENT_SECRET_PATH
--service-account	Service account path	From SERVICE_ACCOUNT_PATH

Analyze Images Only

Extract layout JSON without creating slides:

uv run images2slides analyze \
  --image infographic.png \
  --output layouts/

Creates infographic_layout.json with extracted text and bounding boxes.

Other Commands

# Validate a layout file
uv run images2slides validate --layout layout.json

# Post-process a layout (clean up whitespace, clamp bounds)
uv run images2slides postprocess --layout raw.json --output clean.json

# Build slides from pre-existing layout files
uv run images2slides create \
  --layout slide1.json \
  --layout slide2.json \
  --title "My Deck"

---

Programmatic Usage

from images2slides.vlm import VLMConfig, extract_layout_from_image
from images2slides.auth import get_slides_service_oauth
from images2slides.build_slide import build_presentation, SlideInput
from images2slides.postprocess import postprocess_layout

# Configure VLM
config = VLMConfig(provider="google", model="gemini-3-pro-preview")

# Extract layouts from images
layouts = []
for image_path in ["slide1.png", "slide2.png"]:
    layout = extract_layout_from_image(image_path, config)
    layout = postprocess_layout(layout)
    layouts.append(layout)

# Authenticate with Google Slides
service = get_slides_service_oauth("secrets/client_secret.json")

# Build presentation
slide_inputs = [SlideInput(layout=layout) for layout in layouts]
result = build_presentation(
    service=service,
    slides=slide_inputs,
    title="My Presentation",
)

print(f"Created: {result.presentation_url}")

---

Configuration Reference

Environment Variables

All variables can be set in .env or as environment variables:

Variable	Description
VLM_PROVIDER	VLM provider: google, openai, anthropic, openrouter
VLM_MODEL	Model name (optional, uses provider default)
GOOGLE_API_KEY	Google AI API key
OPENAI_API_KEY	OpenAI API key
ANTHROPIC_API_KEY	Anthropic API key
OPENROUTER_API_KEY	OpenRouter API key
CLIENT_SECRET_PATH	Path to OAuth client secret JSON
SERVICE_ACCOUNT_PATH	Path to service account JSON
GCS_BUCKET	GCS bucket for image uploads (optional)

Example .env File

# VLM Configuration
VLM_PROVIDER=google
GOOGLE_API_KEY=AIza...

# Google Slides Authentication
CLIENT_SECRET_PATH=secrets/client_secret.json

---

Development

Setup

uv sync --dev

Testing

uv run pytest

Code Quality

uv run ruff check images2slides cli
uv run black images2slides cli

---

Troubleshooting

"API key not found"

Make sure you have:

1. Created a .env file from .env.example
2. Set the API key for your chosen provider
3. The .env file is in the current directory or a parent directory

"Must provide --client-secret or --service-account"

You need Google Slides API credentials. See Configure Google Slides API Access.

OAuth browser doesn't open

If running on a headless server, use a service account instead of OAuth.

Service account can't access presentation

Service accounts create files in their own Drive. Either:

• Check the presentation URL in the output and open it
• Share the presentation with your email
• Use OAuth instead for personal use

---

License

MIT