InnoCaptcha 2.2.1

A professional, pluggable CAPTCHA library with image, math, and custom challenge types, token-based security, and multiple storage backends.

InnoCaptcha

A pluggable Python CAPTCHA library supporting image-based text challenges, arithmetic challenges, audio challenges, token-based security, and multiple storage backends.

PyPI · GitHub · Issues · Discussions

---

Table of Contents

• Installation
• Quick Start
  • Text CAPTCHA
  • Math CAPTCHA
  • Audio CAPTCHA
  • Image CAPTCHA
  • Command-Line Interface
• API Reference
• Requirements
• License

---

Installation

pip install InnoCaptcha

---

Quick Start

1. Text CAPTCHA

Generates an image-based CAPTCHA with configurable text, colors, and dimensions. All images include random distortions and anti-aliasing.

from InnoCaptcha.text import TextCaptcha

# Basic usage
captcha = TextCaptcha()
captcha.create("abs")
print(captcha.verify("abs"))    # True
captcha.save(r"captcha.png")

# Custom dimensions and colors
captcha = TextCaptcha(
    width=350,
    height=100,
    color=(255, 137, 6),
    background=(15, 14, 23)
)
captcha.create()
print(captcha.verify('Answer'))
captcha.save(r"captcha.jpg")

Constructor Parameters

Parameter	Type	Default	Description
width	int or None	300	Image width in pixels.
height	int or None	80	Image height in pixels.
color	tuple[int, int, int]	Random color	Foreground (text) color in RGB.
background	tuple[int, int, int]	Random color	Background color in RGB.

create(chars: str) — The text string to render in the CAPTCHA image. Optional.

save() Parameters

Parameter	Type	Default	Description
path	str	'captcha.png'	Full file path or file name to write the image.

Notes:

• Uses secrets for cryptographically strong randomness.
• Rendering can be tuned via module-level constants such as CHARACTER_OFFSET_DX and WORD_SPACE_PROBABILITY.

---

2. Math CAPTCHA

Generates arithmetic challenges (addition, subtraction, multiplication, division). All results are integers — the problem regenerates automatically if division would produce a fraction. Can output as plain text or an image.

from InnoCaptcha.math import MathCaptcha

# Text output (default)
challenge = MathCaptcha()
print(challenge.get_question())  # e.g., "7 + 3 = ?"
print(challenge.answer)          # e.g., "10"

# Image output
image_challenge = MathCaptcha(output="image")
image_challenge.get_question().show() # Returns a PIL.Image

print(challenge.verify(10))      # True
print(challenge.verify("10"))    # True — string input accepted

---

3. Audio CAPTCHA

Generates a spoken character sequence as a WAV file. Each character is spliced from pre-recorded audio samples stored in the data/ directory, with per-character noise injection, randomized playback speed, and a low-pass filter applied to the combined output. Challenge state is persisted in SQLite with a 5-minute expiry and a 5-attempt limit.

from InnoCaptcha.audio import AudioCaptcha

# Basic usage
captcha = AudioCaptcha()
captcha.create("ab3")
captcha.save("captcha.wav")

print(captcha.verify("ab3"))    # True
print(captcha.verify("wrong"))  # False

create(chars: str) — Accepts up to 6 characters. Each character must have a corresponding <char>.wav file in the data/ directory. Raises FileNotFoundError if any file is missing.

save() Parameters

Parameter	Type	Default	Description
path	str	—	Full file path to write the output WAV file.

verify(user_input: str) -> bool or str

Return value	Condition
True	Input matches the stored answer (case-insensitive).
False	Input does not match; attempt counter incremented.
str (error message)	Captcha expired or maximum attempts (5) reached.

Notes:

• Uses secrets for randomness in noise generation and speed variation.
• A background thread runs on instantiation to purge expired records from the database.
• Output is a 44100 Hz, 16-bit, mono WAV file.

---

4. Image CAPTCHA

Presents a 3×3 grid overlay on a randomly selected image. The user identifies which grid cells contain the target object. Detection is performed using YOLOv11n — cells are marked correct if any detected bounding box overlaps them. Challenge state is persisted in SQLite with a 5-minute expiry and a 6-attempt limit.

from InnoCaptcha.image import ImageCaptcha

captcha = ImageCaptcha()
captcha.create()
print(captcha.verify())  # True, False, or str on expiry/lockout

create() — Loads a random image from the dataset, runs YOLO inference to locate objects, and draws a 3×3 blue grid over the result. Must be called before verify().

verify() -> bool or str — Displays the gridded image and prompts the user to enter the cell numbers (1–9, comma-separated) containing the detected object. Returns True only if the submitted cells exactly match all cells that overlap a detected bounding box.

verify(user_input: str) -> bool or str

Return value	Condition
True	Submitted cells exactly match all detected object cells.
False	Input does not match; attempt counter incremented.
str (error message)	Captcha expired or maximum attempts (6) reached.

Notes:

• Grid numbering is row-major: 1–3 top row, 4–6 middle row, 7–9 bottom row.
• Image dataset must be structured as data/images/<class>/<filename>.
• Uses secrets for random image selection.
• A background thread runs on instantiation to purge expired records from the database.

---

5. Command-Line Interface

# Display the installed version
InnoCaptcha --version

# Upgrade to the latest release on PyPI
InnoCaptcha --upgrade

---

API Reference

TextCaptcha

Method / Attribute	Description
create(chars: str)	Renders the given string into a distorted CAPTCHA image.
verify(input: str)	Returns True if input matches the generated text.
save(path)	Writes the image to the specified file path.

MathCaptcha

Method / Attribute	Description
get_question() -> str | PIL.Image.Image	Returns the challenge string (e.g. "7 + 3 = ?") or a rendered PIL Image if output="image".
answer: str	The correct string integer answer to the current challenge.
verify(input) -> bool	Returns True if input equals the answer.

AudioCaptcha

Method / Attribute	Description
create(chars: str)	Builds the audio challenge from up to 6 characters using per-character WAV files.
verify(input: str)	Returns True on match, False on mismatch, or a str on expiry/lockout.
save(path: str)	Writes the generated audio to a 44100 Hz 16-bit mono WAV file.
id: str	The hex token identifying this challenge in the database.
audio: np.ndarray	Raw float32 audio samples; None until create() is called.

ImageCaptcha

Method / Attribute	Description
create()	Runs YOLO detection on a random dataset image and overlays a 3×3 grid.
verify() -> bool or str	Displays the image, accepts grid input, returns True on exact match.
id: str	The hex token identifying this challenge in the database.
image_class: str	The randomly selected object class for the current challenge.
annotation_coordinates	List of (x1, y1, x2, y2) bounding boxes from YOLO inference.

---

Requirements

• Python 3.9 or later
• Pillow >= 10.0.0
• numpy
• scipy
• ultralytics
• opencv-python

---

License

MIT — InnoSoft Company