Changelog

July 27, 2026

Fixed: Image → DXF conversions now work

Converting images (PNG, TIFF, JPG) to DXF vector drawings was failing on our server because the vectorizer needed a graphical display. It now runs with a virtual display, so image-to-DXF conversions complete reliably.

July 23, 2026

Fixed: RAR archive conversion fixed

Converting RAR archives to ZIP was failing for modern RAR5 files. Installed a full RAR extractor, so RAR-to-ZIP now works across all RAR versions.

July 21, 2026

Fixed: Mislabeled files now convert instead of failing

Audio and image files uploaded with the wrong extension — e.g. an MP3 saved as .flac, an M4A saved as .mp3, or a PNG saved as .jpeg — are now detected by their real content and converted anyway, instead of being rejected with a “not a valid file” error.

Fixed: Deeply-nested JSON no longer crashes conversion

Converting a JSON file with very deep nesting could fail with an internal recursion error. Raised the parser limit and made malformed-JSON errors report cleanly.

Fixed: EPUB → MOBI conversion restored

EPUB-to-MOBI (Kindle) conversion had been failing because the underlying tool was unavailable. Switched to Calibre, so EPUB files convert to MOBI again.

Fixed: Several failing conversions now work again

A group of conversions that were failing on our server are working again: JSON → Excel (XLSX), image → SVG, PDF ↔ EPUB, 3D models (FBX/STL/OBJ), and 7z archive extraction. The required tools were missing and have now been installed.

April 29, 2026

Fixed: MCP endpoint no longer triggers PHP warnings

JSON-RPC POST /mcp requests were generating a PHP Warning: Undefined array key "action" in kernel.php because the routing code assumed every POST has a form-encoded action field. Added an isset() guard so MCP and API JSON payloads pass through cleanly.

New: Algolia-powered site search

The header search field is now connected to Algolia. Type 2+ characters to see instant results across all converters, tools, and guides (767 pages indexed in 7 languages). Keyboard navigation with ↑/↓/Enter/Escape and ⌘K shortcut to focus. Results show type badges (Converter, Tool, Guide) and category tags.

New: Redesigned site header (v2)

Simplified navigation: Logo, Converters mega-menu (combined Convert + Tools), search bar, theme switcher, language switcher. On mobile: clean hamburger drawer with search and category links. The old 9-slot header replaced with 5 focused elements.

New: Guides section on homepage

A new “Guides & Tutorials” section appears on the homepage with 6 latest articles from diverse categories (AI Image, AI Audio, Video, Image, etc.). Fully translated across all 7 languages.

Improved: Code blocks now use dark theme in both light and dark mode

Inline code snippets and syntax-highlighted code blocks now consistently use a dark background (Catppuccin Mocha) regardless of the site theme. This matches the industry standard (GitHub, VS Code) and improves readability in light mode.

April 22, 2026

Fixed: Screaming Frog audit: 300+ internal 404s cleaned across converter cross-links, CTAs, and state picker

SEO crawl surfaced seven distinct 4xx causes. Cleared all of them:

• Dead “related converter” links — 15 slugs (pdf-to-tiff, mp3-to-mp4, video-to-gif, rotate-video, xlsx-to-json, csv-to-pdf, csv-to-json, pdf-to-mobi, flac-to-ogg, mobi-to-pdf, pdf-to-word, avi-to-mp3, add-border, mobi-to-epub, png-to-tiff) referenced in src/data/pages/<parent>/{en,de,fr,es,it,pt,nl}.php but never built. Removed 123 stale array entries across 7 languages.
• Remaining double-prefixed CTAs (/de/de/speech-to-text, /de/de/resize-image, etc.) — the 2026-04-17 sweep missed 12 templates that used a ternary variant $_langPrefix ? "/{$_langPrefix}" . $_p[...] : $_p[...]. Stripped the legacy prefix logic from resize-image/{facebook,pinterest,youtube,linkedin}, speech-to-text/{transcribe-audio,audio-to-text,transcribe-interview,transcribe-podcast,video-to-subtitles}, and vocal-remover/{isolate-vocals,karaoke-maker,separate-drums}.
• State-picker 404s on /random-phone-number — the landing page linked to 50 per-state URLs (/random-phone-number/alabama etc.) that were never routed. Converted the block into interactive buttons that set the dropdown, trigger generation, and scroll to the output — no broken URLs, same UX.
• API error docs pointed to /api/reference/convert (non-existent sub-page). Fixed all 7 language files to use the #convert anchor on the existing /api/reference page.
• Escaped-quote hrefs in pdf-page-remover/de.php — href=\"/de/merge-pdf\" inside a single-quoted PHP string emitted literal backslash-quotes, producing unusable hrefs like /de/\"/de/merge-pdf\". Replaced with proper double quotes.
• Cross-language body link to untranslated article — m4a-to-wav/wav-quality-settings linked to /xx/m4a-to-wav/soxr-resampler in all 6 non-English languages, but the article only lives under mp3-to-wav in English. Stripped the broken link wrapper, kept the phrase inline.

Net result: ~300 internal 404 URLs cleared with minimal template changes; no new converter pages built (links simply removed where the target didn't exist). Cloudflare /cdn-cgi/l/email-protection 404s (1431 inlinks) are a CF email-obfuscation artefact and need to be ignored at the crawler level, not fixed in code.