Ink Recipe Software

January 12

Does anyone know of any AI program (or any other software) which is able to analyze visual input (like images of ink swatches), text from the Internet (properties of inks from ink reviews and ink brand websites, like saturation, vibrancy, viscosity etc.) and also has some knowledge of ink mixing; so that one may give it a list of available inks with themselves, along with a prompt to give an idea about the type of color and property they would like to create. Based on this the tool might suggest ink recipes based on the capabilities described above. I am new to ink mixing and I have almost 70 inks, such a tool would be of immense help, I feel with the technology level in the world right now, this should be possible. If any dedicated software for this purpose is not available, do anyone of you use something else for a similar purpose in a slightly different manner? Anything would be a great help. Searching for ink recipes on the Internet by myself is very time consuming and also less accurate.

January 12

I asked Chat GPT. Here is the answer.

Existing Technologies That Can Help

Visual Analysis of Ink Swatches:
- OpenCV (Python Library): A powerful library for image processing. It can extract color data (RGB/HEX values) from ink swatch images.
- Colormind API: Generates palettes and suggests similar colors.
- DeepAI: Has pre-built models for color detection and analysis.
Text Analysis for Ink Reviews:
- GPT-Based Models (like ChatGPT with plugins): Excellent at extracting specific properties like saturation, viscosity, or vibrancy from text reviews or brand descriptions.
- Beautiful Soup + Scrapy: Python tools for web scraping to gather data from ink reviews and brand websites.
- Sentiment Analysis: Tools like TextBlob or VADER can interpret qualitative text (e.g., “rich saturation”) into measurable attributes.
Ink Mixing Knowledge and Simulation:
- Color Science Libraries:
  - Python libraries like colorspacious or colour-science can simulate color mixing using mathematical models.
- Mixing Simulators:
  - There are tools like InkMixTool which offer basic functionalities to simulate ink mixing visually, though they may not integrate with AI for other features.
Combining All in a Custom Program:
- You could integrate these components into a single program using a platform like:
  - Streamlit (for an interactive UI).
  - PyTorch or TensorFlow (to train any custom AI models for more nuanced recommendations).
  - Node.js or Python Flask (for backend logic).

Your AI Tool’s Workflow

Input Phase:
- Upload photos of ink swatches.
- Provide text reviews or let the program scrape them automatically.
- List available inks you own with key details.
Processing Phase:
- Analyze uploaded ink swatches for color properties (e.g., hue, brightness).
- Scrape and parse text to extract ink characteristics like flow, saturation, and drying time.
- Simulate ink mixing using AI to predict new colors and properties.
Output Phase:
- Generate a proposed recipe for the desired color and properties.
- Provide a preview of the color, along with estimated properties.

Possible Challenges

Precision in Mixing: Achieving real-world accuracy in predicting ink properties requires a detailed database of inks, and AI would need to generalize from known examples.
Data Availability: Ink reviews and properties might not always be consistently formatted or available online.
Complexity: Simulating the physical and chemical interaction of inks accurately could be challenging without detailed data.

What to Do Next?

Build It Yourself:
- Start by combining OpenCV for image analysis and GPT or scraping tools for text extraction. Add a simple mixing simulator with color science libraries.
Explore Existing Alternatives:
- Check tools like InkMixTool and see if you can complement them with custom scripts.
Collaborate:
- If you’re not into programming, consider collaborating with a developer to bring your idea to life.

January 12

49 minutes ago, vorpal said:

I asked Chat GPT. Here is the answer.

Existing Technologies That Can Help

Visual Analysis of Ink Swatches:

OpenCV (Python Library): A powerful library for image processing. It can extract color data (RGB/HEX values) from ink swatch images.

Colormind API: Generates palettes and suggests similar colors.

DeepAI: Has pre-built models for color detection and analysis.

Text Analysis for Ink Reviews:

GPT-Based Models (like ChatGPT with plugins): Excellent at extracting specific properties like saturation, viscosity, or vibrancy from text reviews or brand descriptions.

Beautiful Soup + Scrapy: Python tools for web scraping to gather data from ink reviews and brand websites.

Sentiment Analysis: Tools like TextBlob or VADER can interpret qualitative text (e.g., “rich saturation”) into measurable attributes.

Ink Mixing Knowledge and Simulation:

Color Science Libraries:

Python libraries like colorspacious or colour-science can simulate color mixing using mathematical models.

Mixing Simulators:

There are tools like InkMixTool which offer basic functionalities to simulate ink mixing visually, though they may not integrate with AI for other features.

Combining All in a Custom Program:

You could integrate these components into a single program using a platform like:

Streamlit (for an interactive UI).

PyTorch or TensorFlow (to train any custom AI models for more nuanced recommendations).

Node.js or Python Flask (for backend logic).

Your AI Tool’s Workflow

Input Phase:

Upload photos of ink swatches.

Provide text reviews or let the program scrape them automatically.

List available inks you own with key details.

Processing Phase:

Analyze uploaded ink swatches for color properties (e.g., hue, brightness).

Scrape and parse text to extract ink characteristics like flow, saturation, and drying time.

Simulate ink mixing using AI to predict new colors and properties.

Output Phase:

Generate a proposed recipe for the desired color and properties.

Provide a preview of the color, along with estimated properties.

Possible Challenges

Precision in Mixing: Achieving real-world accuracy in predicting ink properties requires a detailed database of inks, and AI would need to generalize from known examples.

Data Availability: Ink reviews and properties might not always be consistently formatted or available online.

Complexity: Simulating the physical and chemical interaction of inks accurately could be challenging without detailed data.

What to Do Next?

Build It Yourself:

Start by combining OpenCV for image analysis and GPT or scraping tools for text extraction. Add a simple mixing simulator with color science libraries.

Explore Existing Alternatives:

Check tools like InkMixTool and see if you can complement them with custom scripts.

Collaborate:

If you’re not into programming, consider collaborating with a developer to bring your idea to life.

Thanks for the information sir, I too searched ChatGPT, but all the methods are too technical and require either considerable programming skills, or navigating and using different softwares in series. Frankly, I was hoping for an easier solution out, anyways the answers are never going to be perfect, I would still have to experiment, albeit more intelligently. These solutions seem like taking too much trouble, for a small gain (at least in the initial days - learning curve etc.) Also, some tools mentioned by GPT just don't work anymore, like I am not able to find this InkMixTool.

2025-01-29T04:04:05Z

In addition to the knowledge of ink(-structure)s and their chemical and physical properties, you also need the concentration of the inks and which additives the final inks contain and many other things. You change the concentration and it behaves totally different. Same is valid for the pH.

The experience of ink mixing is very technical and must be acquired. Ink companies have sometimes decades of experience. Certainly, this experience is not written down anywhere where it could have been picked up by AI training data sets.

If you mix some inks without knowing what you are mixing, chances are that you precipitate the ink as a pigment. And, not always, you can see that pigment without filtration or some analytical methods (requiring apparatus).

My advice: start very small with paper-chromatography to see which components your inks contain. Very few inks contain only one ink.

Example: https://www.instagram.com/p/CFjGdmODogi/

This shows only the colored components, if you use an UV-lamp, you can often (not always) see if there are non-colored components. If you see them, you still cannot say what they are. If you don't have a real chemical laboratory, you cannot reverse-engineer an ink. That is for sure.

Collect property data from the internet.

The 2017/12 #44 issue of the Japanese pen magazine 趣味の文具箱 investigated 144 inks with paper chromatography.

You can buy it used at Amazon: https://www.amazon.co.jp/趣味の文具箱vol-44-エイムック-3909-趣味の文具箱編集部/dp/4777949257

2025-01-29T04:35:20Z

A simple to use program would be a boon to our hobby. I enjoy mixing inks. I only do small amounts and I mostly want waterproof properties in my regular inks. So far, so good, but it would be nice to know ahead of time what I would get.

2025-01-29T06:05:27Z

1 hour ago, Doc Dan said:

mostly want waterproof properties in my regular inks.

How do you want to achieve this?

Real waterproof inks require pigments.

Waterresistant inks require the addition of a binder which traps the ink when dried.

Of course, there is also the method using iron-gall inks.

2025-01-29T15:46:35Z

I just mix the inks. That's it. I get the color I want. Sometimes, if I wash the page one color sort of washes off and the waterproof one stays, though tinted. Sometimes, after they sit for a while, if I wash a page neither washes off. Basically, I experiment.