Devlog 03

The next question was, “Can it even open the kind of images I use?” I used Pillow because it made loading and converting images much less scary.

---

What I did

• Added a file picker for image files.
• Loaded images with Pillow and converted them to RGB.
• Put the image-loading work into image_processor.py instead of mixing it into the GUI.

---

Current result

• Users can choose an image from their computer.
• The program has a separate place for image-related code.

Devlog 03

The next question was, “Can it even open the kind of images I use?” I used Pillow because it made loading and converting images much less scary.

---

What I did

• Added a file picker for image files.
• Loaded images with Pillow and converted them to RGB.
• Put the image-loading work into image_processor.py instead of mixing it into the GUI.

---

Current result

• Users can choose an image from their computer.
• The program has a separate place for image-related code.

Devlog 02

I started with the least exciting but most reassuring thing: making an actual window appear. Seeing a blank app on screen made the project feel real.

---

What I did

• Set up main.py to start the program.
• Created a basic Tkinter app shell.
• Added a title, a simple layout, and a place for image controls.

---

Current result

• The app launches without needing a terminal-focused workflow.
• There is a foundation to build the interface onto.

Devlog 02

I started with the least exciting but most reassuring thing: making an actual window appear. Seeing a blank app on screen made the project feel real.

---

What I did

• Set up main.py to start the program.
• Created a basic Tkinter app shell.
• Added a title, a simple layout, and a place for image controls.

---

Current result

• The app launches without needing a terminal-focused workflow.
• There is a foundation to build the interface onto.

Devlog 01 — The idea

So I was thinking: what if I could give a program a picture and have it draw it with my mouse? It sounded simple at first, but then I realised the computer needs a list of tiny movements, not just an image.

What I did

• Defined the goal: load an image, turn it into strokes, and reproduce them on screen.
• Chose Python because I already knew enough to experiment quickly.
• Picked Tkinter for the desktop window and wrote down the main parts the project would need.

Current result

• A clear plan instead of just “make a drawing bot.”
• The first rough project folder and entry-point idea.

SES-3

This session focuses on bug fixing , CI/CD improvement and some research about new features.

New Features Research

I realized for the past 2 weeks I have not implemented new features, and only fixed bugs and did improvements. I researched a little and looked into other QR libraries and looked for the stuff that the other libraries lack. I decided on some new features like Kanji Mode, there are few but I wont mention them here. (research time isnt logged cause I forgot but it was like an hour)

Bug Fixing

Did some bug fixing with the logo embedding QR code again. Its the only QR code type that fails the most. This time I guarantee it wont fail again.

CI/CD Improvements

Finally I realized for my test workflows I use pyzbar which is a qr code decoding library. But it does not support microqr decoding. So I had to switch to another libray called zxingcpp which does support what pyzbar has with microqr decoding support as well. I took some help from some people online regarding the tests and they help me make some test scripts. I broke down the main test script into smaller parts so Its easier for me and the readers.

Conclusion

Thats it for this session. Its like 95% completed for shipping and logged 50 hours as well. Just some final new features (mentioned above) and some documentation work and It v2 will be released. Thank you for reading the entire thing. I hope you’ll like it and find BetterQR useful. New features / suggestions / complaints are accepted in the replies via a DM on slack. I need some honest feedback.

Initial setup

This is the first devlog for TCast. So here’s what I did.

---

Deciding the structure

TCast should run pretty fast. It has to spawn a new shell , record the inputs, and render them again. So choosing a good language is compulsory. I went with javascript ig.

---

Folders

Setup the initial folders and files needed for the app to work.

---

Research

I researched about other simillar apps to this.

Fixing Changes

Got a review saying the readme is too Ai. So I just wrote a readme from scratch. Cut down the Ai stuff and followed the guide and made this. Also changed the project banner to the terminal preview command cause the reviewer said so.

Final Devlog of v2.0.0 (hopefully)

This is the final devlog that I’ll be posting for v2.0.0 of betterqr. Hopefully you guys will give me some more ideas and features for a future version.

What I did.

Repository Cleanup

• Did a clean full push to the /src and have no uncommited changes.
• Pushed the webiste thats going be hosted on Github Pages.
• Ran some tests.
• Merged the feat/v2.0.0 into main branch.

PyPi Configs

• Changes some paths in links to adjust with the branch merge.
• Modified some automations.

Docs

• Cooked up a release log.
• Did some changes to the README.md and DOCUMENTATION.md

Whats Next?

This what I have been working for around 50 hours. Lets See what happens. Im plannning to maintain this as much as I can.

Grinding

I forgot to devlog so this will be a big one lol. Im currently doing exams so I dont really have alot of time to think through new stuff. I mostly did some bug fixes , optimisation and working on the landing page.

Bug Fixes

• Fixed some bugs related to the PDF output where sometime QR codes overlap and make an error.
• Fixed some bugs related to the command line.
• The frame generation does not work. Im actively working on it but Im not able to fix it yet.

Optimisation

• Have been cutting down extra generation time with unnecssary checks.
• Have imporved alot in the generation time.
• This took the longest.

Docs

• Edited the documentation.md with the new data.
• Edited the readme.md with new stuff.
• Working on a landing page for BetterQR.

I also researched a little bit and asked for peoples opinion on this. Selected around 20 people to test this and give their honest review. Most feedbacks were great. Some requested for features and some found bugs.

Session 5

More Benchmarking

I have benchmarked more and repeatedly to get the accurate data. I used automated and manual tests. I found some people online who were down to help me benchmark it. For now the data isnt fully presentable. Will include in the next devlog or in the readme itself.

New Brand

I have decided on updating our logo. I designed a logo on Canva. I’ll probably use it as the official logo. I went for the purplish theme cause why not. The logo has a futuristic vibe to it ig.

Better Landing Pages.

Im currently working on a landing page for betterQR. The old one is fine but I didnt update it accordingly. So Im going to make a entirely new one. I have been looking for some inspo as well.

Bug Fixing

Did some little bug fixes and code cleanup

The tables does not work on the stardance devlogs, idk why look at the attached image for the formatted table

Benchmarking

This session focuses on benchmarking BetterQR with other popular libraries like segno and qrcode.

How it Compares

This is a rough comparison between BetterQR, qrcode, and segno.

Feature BetterQR qrcode segno Micro QR ✅ M1–M4 ❌ ✅ Kanji Mode ✅ ❌ ✅ Structured Append ✅ ❌ ✅ Logo Embedding ✅ Built-in ❌ (Manual PIL) ❌ (Manual) Module Shapes ✅ 8 Shapes ⚠️ Limited ❌ Gradients ✅ ❌ ❌ Animated GIF ✅ 10 Effects ❌ ❌ Frames & Labels ✅ Built-in ❌ ❌ Structured Data Helpers (WiFi, vCard, …) ✅ ❌ ⚠️ Via plugin

BetterQR’s primary aim is not to aggressively beat other libraries, but to provide a highly capable, reliable, and user-friendly alternative packed with creative options out of the box.

---

Performance Metrics

Payload Performance Benchmarks (2.0.0-rc)

All times are in milliseconds (ms). Lower is better.

| Payload Category | Size / Type | BetterQR | Segno | Standard qrcode | BetterQR Status |

| :— | :— | :—: | :—: | :—: | :— |
| URL | Tiny / Short | 6.88 | 7.19 | 5.90 | 🟢 Beats Segno / Competitive |
| Medium | Standard | 22.44 | 21.89 | 20.03 | 🟡 Ties Segno (~2.5% diff) |
| Large | V40 / Dense | 106.63 | 98.94 | 98.96 | 🔴 Behind by ~8% |
| Very Large | Max Capacity | 209.00 | 197.00 | 204.00 | 🟡 Highly Competitive (~6% diff) |
| SVG (Large) | Vector | 86.63 | ~0.56 | — | 🟡 Meaningful fix (was 104.66) |
| SVG (Very Large)| Vector | 174.19 | ~1.50 | — | 🟡 Meaningful fix (was 211.84) |

Known Limitations / Project Honesty

To build clear engineering trust, here are the current structural limitations of BetterQR compared to older ecosystem veterans:

• Dependency Overhead: Depends heavily on Pillow for its extensive visual layouts and shapes, unlike segno, which is zero-dependency.
• Large Matrix Vectorization: Pure PNG generation lags behind by ~6-8% on maximum matrix dimensions (like V40) due to mask scoring loops running entirely in pure Python.
• Vector Vectorization Limits: While our SVG generation times dropped significantly in the latest release, pure string-building layout engines like segno still massively outperform us on massive vector layouts.
• Format Boundaries: The legacy EPS writer does not support logo embedding.
• Spec Edge Cases: Kanji mode and Structured Append are fully validated for standard QR codes, but are not yet implemented for Micro QR variants. rMQR (Rectangular Micro QR) is currently unsupported.

---

The tables does not work on the stardance devlogs, idk why look at the attached image for the formatted table

Benchmarking

This session focuses on benchmarking BetterQR with other popular libraries like segno and qrcode.

How it Compares

This is a rough comparison between BetterQR, qrcode, and segno.

Feature BetterQR qrcode segno Micro QR ✅ M1–M4 ❌ ✅ Kanji Mode ✅ ❌ ✅ Structured Append ✅ ❌ ✅ Logo Embedding ✅ Built-in ❌ (Manual PIL) ❌ (Manual) Module Shapes ✅ 8 Shapes ⚠️ Limited ❌ Gradients ✅ ❌ ❌ Animated GIF ✅ 10 Effects ❌ ❌ Frames & Labels ✅ Built-in ❌ ❌ Structured Data Helpers (WiFi, vCard, …) ✅ ❌ ⚠️ Via plugin

BetterQR’s primary aim is not to aggressively beat other libraries, but to provide a highly capable, reliable, and user-friendly alternative packed with creative options out of the box.

---

Performance Metrics

Payload Performance Benchmarks (2.0.0-rc)

All times are in milliseconds (ms). Lower is better.

| Payload Category | Size / Type | BetterQR | Segno | Standard qrcode | BetterQR Status |

| :— | :— | :—: | :—: | :—: | :— |
| URL | Tiny / Short | 6.88 | 7.19 | 5.90 | 🟢 Beats Segno / Competitive |
| Medium | Standard | 22.44 | 21.89 | 20.03 | 🟡 Ties Segno (~2.5% diff) |
| Large | V40 / Dense | 106.63 | 98.94 | 98.96 | 🔴 Behind by ~8% |
| Very Large | Max Capacity | 209.00 | 197.00 | 204.00 | 🟡 Highly Competitive (~6% diff) |
| SVG (Large) | Vector | 86.63 | ~0.56 | — | 🟡 Meaningful fix (was 104.66) |
| SVG (Very Large)| Vector | 174.19 | ~1.50 | — | 🟡 Meaningful fix (was 211.84) |

Known Limitations / Project Honesty

To build clear engineering trust, here are the current structural limitations of BetterQR compared to older ecosystem veterans:

• Dependency Overhead: Depends heavily on Pillow for its extensive visual layouts and shapes, unlike segno, which is zero-dependency.
• Large Matrix Vectorization: Pure PNG generation lags behind by ~6-8% on maximum matrix dimensions (like V40) due to mask scoring loops running entirely in pure Python.
• Vector Vectorization Limits: While our SVG generation times dropped significantly in the latest release, pure string-building layout engines like segno still massively outperform us on massive vector layouts.
• Format Boundaries: The legacy EPS writer does not support logo embedding.
• Spec Edge Cases: Kanji mode and Structured Append are fully validated for standard QR codes, but are not yet implemented for Micro QR variants. rMQR (Rectangular Micro QR) is currently unsupported.

---

SES-3

This session focuses on bug fixing , CI/CD improvement and some research about new features.

New Features Research

I realized for the past 2 weeks I have not implemented new features, and only fixed bugs and did improvements. I researched a little and looked into other QR libraries and looked for the stuff that the other libraries lack. I decided on some new features like Kanji Mode, there are few but I wont mention them here. (research time isnt logged cause I forgot but it was like an hour)

Bug Fixing

Did some bug fixing with the logo embedding QR code again. Its the only QR code type that fails the most. This time I guarantee it wont fail again.

CI/CD Improvements

Finally I realized for my test workflows I use pyzbar which is a qr code decoding library. But it does not support microqr decoding. So I had to switch to another libray called zxingcpp which does support what pyzbar has with microqr decoding support as well. I took some help from some people online regarding the tests and they help me make some test scripts. I broke down the main test script into smaller parts so Its easier for me and the readers.

Conclusion

Thats it for this session. Its like 95% completed for shipping and logged 50 hours as well. Just some final new features (mentioned above) and some documentation work and It v2 will be released. Thank you for reading the entire thing. I hope you’ll like it and find BetterQR useful. New features / suggestions / complaints are accepted in the replies via a DM on slack. I need some honest feedback.

Bug Patching (SES-2.1)

This will be my last patch of the day.

Bug

• Styled Micro QR codes render wrong: Both the PNG and SVG renderers assume the standard 3-finder layout when deciding what counts as a “finder square” (for shape/color overrides). Micro QR only has one finder (top-left), so modules near the top-right and bottom-left corners get incorrectly forced into squares / finder-color under custom styling.

This isnt fully fixed, Im having some hard time figuring out whats the main core issue. It is somewhat fixed.

Bug Patching (SES-1)

Ran more tests and got feedback from ~15 people. Made them test the commands and try to find bugs. There were many bugs people were available to find. This is session focuses on fixing some of them

Bugs that were fixed

• Micro QR silently corrupts data near capacity limits: In encoder.py, the capacity check treats every Micro QR data codeword as a full 8 bits — but for M1 and M3 symbols, the last data codeword is only a 4-bit nibble per spec. The checker accepts data that’s 4 bits too big, and the extra bits get silently truncated at render time.
• .logo() crashes every time: core.py builds a logo_info dict with a key "padding", but matrix.py leads logo_info["padding_px"] , instant KeyError the moment anyone calls .logo(...).
• Terminal --terminal-style compact output has padding bugs asymmetric left/right padding and a stray extra row, so piping it anywhere that cares about exact dimensions breaks.

There are some more bugs that will be fixed in the next session.

Mega Lock-In

The 9 hours was logged in span of the last week. Didn`t get to devlog due to internet outage here in Sri Lanka.

Bugs Fixed

in core.py

• _hex_to_rgb 3-char hex crash:#000, #FFF, #F00 etc. caused invalid literal for int() with base 16: ''. Added expansion of 3-char to 6-char before parsing

• .logo() crash: 'QR' object has no attribute '_version_size'``logo() tried to compute size_px = int(self._version_size * ratio) but _version_size doesn’t exist on the QR object at call time (the matrix hasn’t been built yet). Fixed to compute size_px lazily inside _get_logo_info() from module_count × box_size × ratio.

• logo(border=True) crash: invalid literal for int() with base 16: 'Tr'The _logo_border attribute stores a bool, but was passed directly as border_color to embed_image() which called _hex_to_rgb(True) → _hex_to_rgb("True") → crash. Fixed to pass "#000000" if self._logo_border else None.

• image_ops.py + animation.py — _hex_to_rgb None guard
  Added if not c: return (0, 0, 0) guard so passing None as a color never crashes.

• The _color() validator didn’t strip surrounding shell-added quotes from color strings like “#FF0000”, causing # to be split off the hex digits. Fixed to strip " and ' before processing.

Changes

• Corrected Micro QR capacity table (all 8 version/ECC combos with accurate numeric/alpha/byte limits)
• Fixed logo() parameter docs (border is bool, not int)- Fixed double ```bash` in installation section
• Corrected SMS encoding protocol note (sms: not smsto:)

Docs

• Added a more detailed info about MicroQR codes to README.md and DOCUMENTATION.md- Made a landing site with help of Claude.

Publishing

Did some final stuff and published to PyPi.

• Add SECURITY.md and `CONTRIBUTION.md
• Fix workflow to auto-publish to PyPi.
• Working a landing website.
• Added a release
• Published to PyPI

Now install via

pip install vokul --upgrade