Changelog
Dated changes to the site, newest first. Every entry describes something a visitor can see or use. For background on a topic, see the guides.
Fewer busy messages when Gemini is slow
When Gemini is slow, the AI tools now switch to a backup model while there is still time. Before, a slow Gemini could use the whole 20 second limit, and you saw the busy message without a backup ever being tried.
The EXIF Viewer now reads WebP photos
A WebP photo used to show "Unknown file format". The EXIF Viewer now shows the camera, date, and GPS tags in a WebP file, and a WebP photo without metadata shows the normal "no metadata" note.
The analyzer keeps your photo when you change mode
Switching between Ask Anything, Profile Picture, Identify, and Calories now keeps your photo and your question. Errors now appear where you can see them: a rejected file shows at the upload area, and a failed run scrolls into view on a phone. The phone action bar now stays with the tool instead of following you down the page.
Every tool now opens ready to use
Tool pages now open on the tool itself: the controls, the input, and the result share one workspace frame on the first screen. The site also moves to one font and one aqua action color, and the pages no longer carry a dark theme that never applied.
The free-use line now waits for a real daily count
When the shared counter is not available, the AI tools no longer show a per-minute number as today's count. The line appears only when the daily device count is available.
A remaining-uses count on every AI tool
The AI tools now show how many free uses are left today, read from the server. The analyzer also enforces the six follow-up turns per image on the server, so a reload no longer resets that count.
BMP and GIF export now works in every browser
The converter writes BMP and GIF itself when the browser canvas cannot. Chromium, Firefox, and WebKit now produce both formats from a real photo. A 1024 by 683 image becomes a 2.1MB BMP or a 220KB GIF.
Clearer messages when an AI tool is busy
A busy free tier, an overloaded model, and a failed upstream request now read differently, and each one says what to do next. The analyzer keeps the reply it already has and offers Try again.
CSV encoding bugs, reproduced and measured
Five CSV files through the CSV Cleaner, three runs each. A 190-byte Windows-1252 file was repaired automatically; a 10,654-byte one with three bad bytes was not, and the retry button fixed it.
Ten JSON errors, with the exact messages
Ten broken inputs through the JSON Formatter, three runs each. The page shows the browser's own message; the guide maps each one to the fix and shows the formatted output.
What four compression qualities actually cost
Three CC0 photos at quality 90, 75, 50, and 25, three runs each. Quality 90 grew every file about 20 percent. Quality 75 cut 17 to 28 percent with little visible loss.
A real EXIF dump, tag by tag
One CC0 photo from a Canon EOS R shows 21 EXIF tags in the viewer. The guide explains each tag, separates the IPTC and XMP rows, and adds five GPS tags to a copy to show the location block.
A real damage test for QR error correction levels
One URL at L, M, Q, and H, then black blocks over 1 to 50 percent of each code. Q and H stopped at 20 percent in the center and 1 percent in the corner, below the quoted 25 and 30 percent.
Real merge and split times for three PDFs
Three NIST PDFs went through PDF Tools three times each. Merge took 1.37 to 1.42 seconds for two files. The 86-page split produced 24.7MB, and a 51MB upload was dropped without a message.
Twelve image prompts, with every answer
Three CC0 photos went through twelve prompts in the AI Image Analyzer. The guide shows the real answers, a stable 7/10 score with reasons that changed, a total that moved by 80 calories, and one prompt that guessed a lake in Colorado.
A clearer AI Image Analyzer layout
The analyzer now opens on a compact hero and a two-pane workbench. The result pane starts on a labeled sample, the follow-up chat sits under the result, and the whole analyzer family uses the site indigo accent.
Identifying plants and objects from photos, with nine runs
Three CC0 photos went through Identify mode three times each. The page shows every run, the stable wrong answer for a monarch butterfly, and a table that separates tested photo advice from general tips.
AI Image Analyzer shows three saved runs
A landscape, a butterfly, and a pasta plate each appear with the output the tool saved on 2026-11-16. The page also explains what image analysis can and cannot do, and the dropzone can load a sample photo.
The older guides now link the source behind each number
Every guide that predates the current series now links its numbers and technical claims to a standard, a vendor document, or a test. A few examples without a source were removed or reworded.
What happens to your photo in an AI tool
A code-level trace of the AI Image Analyzer path, plus a real test that shows when a GPS tag survives the upload and when the Canvas resize removes it.
Token Counter now shows real counts for three texts
The page explains the o200k and cl100k encodings, gives the formulas behind the Claude and Gemini estimates, and lists three measured counts with the limits of each row.
Compress Images now shows a measured run
Three CC0 photos were compressed at 70% quality. The page lists before and after bytes, explains the target-size search, and states which metadata the new file does not keep.
Cursor Rules Generator now shows two real outputs
Two project descriptions were generated into file sets. The page lists the files, byte sizes, and first lines, and explains why the output is a starter rather than a code analysis.
MCP Tool Checker now shows a valid and a broken tool
One clean tool scored 100 and one broken tool scored 89 with four issues. The page explains the rubric, the schema token estimate, and why the score alone is not a pass or fail.
A full photo analysis in one click
The AI Image Analyzer now returns a complete report when the question box is left empty. Result cards also recover when a model reply arrives in an unexpected format, and the page states its free-quota limits.
A clearer 404 page
A missing page now says what happened and links to the AI Image Analyzer, the full tool list, and the guides.
Author bio on About and the homepage
The About page and the homepage now name the developer, describe the work behind the site, and link to the full bio.
Screenshot to Code answers more questions
Four new FAQs cover Figma exports, responsive output, dark mode, and input format. The page also gained an output comparison table.
New design system and rebuilt homepage
The site moved to a flatter visual system with one accent color, and the homepage was rebuilt around the four tool hubs.
Clearer homepage summary and converter pages
The homepage description now matches what people search for. Image conversion pages got titles and descriptions written for each format pair.
Spoke pages added to the navigation
Identify From a Photo, Photo Calorie Estimator, the WCAG Alt Text Checker, and Compress Images are now listed in the header, footer, and hubs.
Sitewide copy pass
Tool pages and guides were rewritten in plain language, with shorter sentences and fewer repeated lines.
More specific converter and FAQ copy
Every image conversion page got its own copy. Eight tool pages gained visible FAQs that match their structured data.
Guides, contact page, and ads.txt
The guides section launched with twelve write-ups, the contact page went live, and ads.txt was published.
Drag and drop on the analyzer
The AI Image Analyzer accepts drag and drop, and the Identify page got its own copy and internal links.
Share a result as an image
The analyzer gained Share PNG and clearer empty states for each mode.
HEIC and AVIF conversion
iPhone HEIC and AVIF files can be converted in the browser. The image converter expanded to 36 direction pages, and the JSON formatter gained highlighting, a tree view, and file upload.
AI Image Analyzer launched
The first version shipped with four modes (Ask, Profile Picture, Identify, Calories) and follow-up chat on the same photo.