imggen - OpenAI Image Generation and OCR CLI

imggen - OpenAI Image Generation and OCR CLI

Generate images from text prompts and extract text from images using OpenAI's APIs.

Overview

imggen is a command-line tool that interfaces with OpenAI's image generation API. It supports multiple models (gpt-image-1, dall-e-3, dall-e-2) and provides options for image size, quality, format, and style.

Prerequisites

• imggen binary installed and available in PATH
• OPENAI_API_KEY environment variable set with a valid OpenAI API key
• Sufficient OpenAI API credits for image generation

Usage

imggen [flags] "prompt"

Available Flags

Model-Specific Parameters

gpt-image-1 (Default, Recommended)

• Sizes: 1024x1024, 1536x1024 (landscape), 1024x1536 (portrait), auto
• Quality: auto, low, medium, high
• Max images: 10 per request
• Supports: Transparent backgrounds, multiple output formats

dall-e-3

• Sizes: 1024x1024, 1024x1792, 1792x1024
• Quality: standard, hd
• Max images: 1 per request
• Supports: Style parameter (vivid/​natural)

dall-e-2

• Sizes: 256x256, 512x512, 1024x1024
• Max images: 10 per request

Instructions

1. Verify OPENAI_API_KEY is set in the environment
2. Construct the imggen command with appropriate flags based on user requirements
3. Execute the command using Bash tool
4. Report the generated filename and any revised prompt returned by the API
5. If the user wants to view the image, use Read tool on the generated file

Output Format

The tool outputs:

• Progress message: "Generating N image(s) with MODEL..."
• Saved filename: "Saved: filename.png"
• Cost information: "Cost: $X.XXXX (N image(s) @ $X.XXXX each)"
• Revised prompt (if returned by API): "Revised prompt: ..."
• Completion message: "Done!"

Generated files are saved to the current working directory with timestamp-based names (e.g., image-20251216-120000.png) unless --output is specified.

Cost Tracking

All image generation costs are automatically logged to ~/​.imggen/​sessions.db. View costs using the cost subcommand:

# View total costs
imggen cost

# View today's costs
imggen cost today

# View this week's costs (last 7 days)
imggen cost week

# View this month's costs (last 30 days)
imggen cost month

# View costs by provider
imggen cost provider

Interactive Mode Cost Commands

In interactive mode (imggen -i), use the cost or $ command:

• cost today - Today's costs
• cost week - This week's costs
• cost month - This month's costs
• cost total - All-time total
• cost provider - Breakdown by provider
• cost session - Current session's costs

Database Management

Manage the SQLite database storing sessions and cost data:

# Reset database (delete all data)
imggen db reset

# Reset with backup of old data
imggen db reset --backup

# Show database location and stats
imggen db info

Examples

Basic image generation

imggen "a sunset over mountains"

High-quality landscape with DALL-E 3

imggen -m dall-e-3 -s 1792x1024 -q hd "panoramic view of a futuristic city"

Multiple images with gpt-image-1

imggen -n 4 -q high "abstract geometric pattern"

Logo with transparent background

imggen -t -f png "minimalist tech company logo, flat design"

Custom output filename

imggen -o hero-image.png "website hero banner with gradient"

Natural style portrait

imggen -m dall-e-3 --style natural "professional headshot, studio lighting"

Multiple prompts via command line

# Generate multiple images with --prompt flag
imggen --prompt "a sunset" --prompt "a cat" --prompt "a dog" -o ./​output

# Short form with parallel processing (3 workers)
imggen -P "sunset" -P "mountains" -P "ocean" -o ./​images -p 3

Batch generation from file

# From a text file (one prompt per line)
imggen batch prompts.txt -o ./​output

# From a JSON file with per-prompt options
imggen batch prompts.json -o ./​output

# With parallel processing
imggen batch prompts.txt -o ./​output -p 3

Multiple Prompts

Generate multiple images from command-line prompts using the --prompt/​-P flag:

imggen --prompt "a sunset over mountains" --prompt "a cat playing piano" -o ./​output

This processes all prompts and saves images to the output directory with indexed filenames:

• 001-a-sunset-over-mountains.png
• 002-a-cat-playing-piano.png

Use --parallel/​-p to control concurrent processing (default: 1 = sequential).

Batch Generation

Generate multiple images from a file of prompts using the batch subcommand:

imggen batch <input-file> [flags]

Input File Formats

Text file (.txt) - One prompt per line (lines starting with # are ignored):

a sunset over mountains
a cat playing piano
abstract geometric art

JSON file (.json) - Array of objects with optional per-prompt settings:

[
  {"prompt": "a sunset over mountains"},
  {"prompt": "a cat playing piano", "model": "dall-e-3", "quality": "hd"},
  {"prompt": "abstract art", "size": "1792x1024"}
]

Batch Flags

Error Handling

Common errors and solutions:

• "API key required": Set OPENAI_API_KEY environment variable
• "invalid size": Use a size supported by the selected model
• "supports maximum N images": Reduce --count value
• "does not support --style": Only dall-e-3 supports style flag
• "does not support --transparent": Only gpt-image-1 supports transparency

Pricing Reference

Costs per image (USD):

gpt-image-1

dall-e-3

dall-e-2

OCR (Optical Character Recognition)

Extract text from images using OpenAI's vision API with optional structured output support.

OCR Usage

imggen ocr <image-path> [flags]

OCR Flags

OCR Models

OCR Examples

Basic text extraction

imggen ocr document.png

Extract from URL

imggen ocr --url https://example.com/​image.png

Save output to file

imggen ocr receipt.jpg -o extracted.txt

Structured output with JSON schema

# Create a schema file (invoice_schema.json):
# {
#   "type": "object",
#   "properties": {
#     "vendor": {"type": "string"},
#     "date": {"type": "string"},
#     "total": {"type": "number"},
#     "items": {
#       "type": "array",
#       "items": {
#         "type": "object",
#         "properties": {
#           "name": {"type": "string"},
#           "price": {"type": "number"}
#         },
#         "required": ["name", "price"],
#         "additionalProperties": false
#       }
#     }
#   },
#   "required": ["vendor", "date", "total"],
#   "additionalProperties": false
# }

imggen ocr receipt.jpg --schema invoice_schema.json -o invoice.json

Auto-suggest a JSON schema

# Analyze image and suggest appropriate schema
imggen ocr document.png --suggest-schema

# Save suggested schema to file
imggen ocr document.png --suggest-schema -o suggested_schema.json

Use higher accuracy model

imggen ocr complex-document.pdf -m gpt-5.2

Custom extraction prompt

imggen ocr business-card.jpg -p "Extract the name, title, email, and phone number"

OCR Structured Output

When using the --schema flag, the output will be structured JSON matching your schema. This is useful for:

• Extracting data from receipts, invoices, forms
• Parsing business cards, ID documents
• Converting tables and structured content to JSON
• Data entry automation

The schema must follow JSON Schema draft-07 format with additionalProperties: false for strict validation.

OCR Tips

1. Use gpt-5-nano for simple text extraction (plain documents, basic receipts)
2. Use gpt-5-mini (default) for most OCR tasks (receipts, business cards, forms)
3. Use gpt-5.2 for complex documents (dense tables, handwriting, multi-language)
4. Suggest schema first if unsure about document structure
5. Custom prompts help when you need specific fields or formatting
6. Supported formats: PNG, JPEG, GIF, WEBP, PDF (first page)