AI and Origami Instructions - Might Things Have Changed?

                 © 2025 by Lindy van der Merwe

AccessOrigami.com

To those of us who, just a year or two ago, eagerly typed our requests for some origami instructions into the edit box of some Large Language Models (top of their game at the time...) and were disappointed to receive some nonsensical steps that did, admittedly, sound quite doable... might things have changed? I will, for now, refrain from an answer and let you decide for yourself.

We were definitely not the only ones to be disappointed. If you have a moment or more, this post about an AI-generated origami craft tutorial for a finger puppet from Reddit sparked some hilarious comments as recently as seven months ago, but it also points out the fact that generative AI is seeping into the very fabric (or is it paper) of our craft. Just one reaction to illustrate:

"goingdownthehill's comment: The most infuriating thing to me about these as a person that folds origami is these things discourage people unfamiliar with the craft even more."

The original post can be found here: https://www.reddit.com/r/mildlyinfuriating/comments/1iibf23/this_ai_generated_origami_tutorial/

If you are interested in the subject of AI and how it relates to origami, the articles by Michal Kosmulski, "A roadmap of using AI in origami," and its follow-up, "Origami AI Revisited," make for interesting re            AI and Origami Instructions - Might Things Have Changed?ading.

Find the articles here: A roadmap of using AI in origami and Origami AI Revisited.

The subject may be approached from many different angles, including issues surrounding the generation Kof origami images and copyright, the use of AI for the identification and classification of models, and, of course, using AI for the teaching of origami. After my initial disappointment with the (totally unusable) folding instructions one would receive from the most capable models at the time, I decided that, given the fact that AI was supposed to get better over time, I would repeat the exercise of asking for instructions for two of the most basic origami bases every so often. As seems to happen with many subjects, different AI models gave different answers, and as we found out in time, the prompt one uses could affect the outcome to a large extent.

In the meantime, apps like Be My Eyes and others have made static and live image description possible, taking advantage of the vision capabilities of some multi-modal AI models. Being able to ask questions about parts of an image or a scene has given us access to information on a whole new level. Even though interaction is often still slow or clunky and sometimes the info is just wrong or a hallucination of the AI, it is likely to improve in the near future. Another aspect that may become important for blind and visually impaired people is the work being done to improve (through fine-tuning and many other techniques) models in specific areas. It is conceivable that models with good reasoning and vision skills may be developed to help us within the areas of science, technology, and maths, making it possible to access visual content like bar and flow charts, graphs of all kinds, and, dare we hope, origami diagrams?

Needless to say, to be on the safer side, if you generate AI instructions you would like to share, please fold the entire model to confirm that the tutorial makes sense, includes all steps, and that the outcome is as expected.

With the above in mind, I am sharing the following two links, but am including the text below for the sake of convenience, so you can decide for yourself if things have, indeed, changed, considering that AI has now been available for around three to four years.

The Square Base and The Waterbomb Base

Here are links to the original posts:

The Square Base - Text Instructions - Google Gemini 2.5 Flash (General Model)

First up is Google Gemini 2.5 Flash (general model/no tools activated) with The Square Base, using the following prompt:

"Hi, I would like some verbal instructions for folding the Square or Preliminary Base in origami. I'd like to start with the vertical folds, then turn over and then the remaining two folds and the collapse. Can you help, including at least two sources? This is for a tutorial I want to put on my origami site for blind people. Thanks."

Except for a small edit in Step 1, the output has been left unchanged.

THE SQUARE BASE - TEXT-BASED INSTRUCTIONS - GOOGLE GEMINI 2.5 FLASH (GENERAL MODEL)

Step 1: The Initial Vertical Folds

Start with your paper in a square shape, with the "right side," or the color/pattern you want to show on the finished model, facing down on the table.

Fold it in half, bringing the bottom edge up to meet the top edge. Make sure the corners and edges are perfectly aligned.

Firmly crease the fold with your fingers. This is the first crease.

Now, unfold the paper completely. You should feel a horizontal crease line running across the middle of the paper.

Next, fold the paper in half again, but this time bring the left edge over to meet the right edge. Align the corners and edges carefully.

Crease this fold firmly.

Unfold the paper completely. You should now feel a horizontal and a vertical crease line that cross in the exact center of the square.

Step 2: Turning the Paper Over

Turn your paper over so the "right side" is now facing up. You should feel the two crease lines you just made. One is horizontal, and one is vertical.

Step 3: The Diagonal Folds

Now, we'll make a different kind of fold. Take the bottom right corner of the paper and bring it up to meet the top left corner. This will make a diagonal fold. Run your fingers along the edges to make sure they are perfectly lined up before you crease.

Crease this fold firmly.

Unfold the paper. You should now feel a diagonal crease from one corner to the opposite corner.

Next, bring the bottom left corner up to meet the top right corner. Align the edges and crease firmly.

Unfold the paper. You should now feel two diagonal creases that cross in the center, forming an "X" shape.

Step 4: The Collapse

You should now have a square with four crease lines that all intersect in the center: one vertical, one horizontal, and two diagonal.

Turn the paper over one more time, so the "right side" is facing down again. The creases will "want" to fold in on themselves.

Place your fingers on the left and right sides of the paper, in the middle of the square.

Gently push in on these sides. As you do, the horizontal and vertical creases should start to push up, and the diagonal creases should start to fold in.

Keep pushing gently until the whole paper collapses into a smaller square. The top and bottom points of the original paper will meet in the middle, and the sides will fold in to create a flat, collapsed shape.

The final result should look like a smaller, flat diamond shape, with four layers of paper at the bottom point and the top point. You can run your fingers along the edges to feel the four sides of the new square. The open, or loose, points are at the bottom, and the closed folds are at the top.

Sources:

  • Origami Resource Centre
  • Origami-Fun.com

The Waterbomb Base - Text Instructions - Chat GPT 5

Our second tutorial is for the Waterbomb Base by Chat GPT 5, also specifying only the sequence of folds, text output, and the use of two sources, with no changes made to the generated output. Note the handy Sequence Summary after the main tutorial, followed by links to all relevant sources.

THE WATERBOMB BASE - Chat GPT 5
  1. Start with a square sheet of paper, colored side up (if using two‑sided paper)—this ensures the colored side will be the exterior later.
  2. Valley‑fold horizontally: fold the top edge down to meet the bottom edge. Crease well then unfold.
  3. Valley‑fold vertically: fold the right edge over to meet the left edge. Crease, then unfold.

At this point, you have created a “+” crease pattern (horizontal and vertical folds).

  1. Flip the paper over, so the white (reverse) side is now facing up.
  2. Valley‑fold along a diagonal: fold one corner to the opposite one (e.g., bottom-left to top-right). Crease, then unfold.
  3. Valley‑fold the other diagonal: fold the remaining two opposite corners together, crease, then unfold.

Now you have created an “X” crease pattern overlaying the “+”, giving you a full crease grid necessary for collapsing.

  1. Collapse into the base (sometimes referred to as a “prayer fold” or “inside reverse folds”):
    • Pinch the left and right edges inward (towards the center), while simultaneously bringing the top and bottom edges as needed to form a triangle shape.
    • This will naturally collapse the paper along the previously made creases, flattening it into a triangular shape with four triangular flaps around the base.

The result is the Waterbomb Base: an isosceles‑right triangle with four flaps—each originating from one side of the square.

Sequence Summary (Plain List)

  • Colored side facing up
  • Fold horizontal (top to bottom), crease, unfold
  • Fold vertical (right to left), crease, unfold
  • Turn paper over (white/blank side up)
  • Fold one diagonal, crease, unfold
  • Fold the other diagonal, crease, unfold
  • Collapse paper into the Waterbomb Base (triangle with four flaps)

Comparison with Trusted Sources

  • Folds.net: Steps 1–5 align with making the horizontal and vertical valley folds, then flipping (step 6) to make diagonal valley‑folds, then collapsing into the base in steps 10–14.
  • Origami‑Resource‑Center: Matches precisely—horizontal + vertical, flip over, diagonal folds, then collapse to finish.

These closely match your requested sequence—with horizontal & vertical first, flip, then diagonal folds, ending in the collapse to form the base.

Additional Clarification from Other Guides

  • Origami.me also instructs: “fold and unfold in half both horizontally and vertically, then turn the paper over, fold/unfold both diagonals, then collapse into the base.”
  • Make‑Origami.com provides the same sequence: “horizontal, vertical, then turn over, diagonal folds, then collapse.”

Sources


Comments

Post a Comment