ConvertekConvertek.tech
Back to Converter
Convertek

Product updates

Changelog

User-facing conversion fixes and product updates. We log issues that affected converting or downloading files so you can see what was resolved.

Last updated: October 29, 2026

Recent releases

Latest
·Ten new conversion guides
  • Ten new in-depth guides were published, covering the conversions people ask about most: PNG to JPG (what transparency costs), JPG to PDF (page size and DPI), PDF to JPG (rendering resolution), PowerPoint to PDF (fonts and lost animation), Excel to CSV (formulas, dates and leading zeros), SVG to PNG (export size and outlined text), MP3 vs WAV vs FLAC (bitrate and lossless archiving), favicons and ICO files (sizes browsers actually request), MOV to MP4 (stream copy versus re-encode) and EPUB vs PDF (reflowable versus fixed layout).
  • Each guide explains the trade-off a conversion makes and links straight to the converter pairs it discusses, so the right format is one click away instead of a guess.
  • The guides index now has a Back to home link at the top, matching the breadcrumb the individual guides already had.
·RAR conversion is fixed
  • RAR files convert again. The server now installs its own rar and unrar binaries on startup and again on the first RAR job, so a deploy or a dyno restart can no longer leave every RAR conversion failing with “RAR packaging is not available on this server”.
  • Creating RAR archives is working: ZIP, 7Z, TAR, TAR.GZ, TAR.BZ2, TAR.XZ, TAR.ZST, GZ, APK and JAR can all be packed to RAR.
  • Opening RAR archives is working too: RAR converts to ZIP, 7Z, TAR, TAR.GZ, TAR.BZ2, TAR.XZ, TAR.ZST, APK, JAR and CBZ.
  • A RAR that genuinely cannot be opened now says why — password-protected, split across several volumes, or corrupt — instead of claiming the server has no RAR support. A toolchain that fails to load also falls back to the built-in extractor instead of aborting the job.
  • The admin overview reports whether rar and unrar are installed on the running dyno, with the resolved path and whether it came from an environment variable, the bundled vendor folder, or PATH.
·Animation, Feather, Stata and P7M fixes
  • FLI and FLC animations now convert instead of failing with “Input image exceeds pixel limit”. The frame header was being read at the wrong offset, so the image size came from unrelated bytes; the first frame is now decoded properly and coloured with the file's palette rather than in greyscale.
  • Feather files saved with compressed record batches (LZ4 or ZSTD, the default in pandas and pyarrow when compression is requested) now convert. Previously every such file failed with “Record batch is compressed but codec not found”.
  • A Stata .dta file that cannot be read now reports the real reason — the release it declares and the byte where parsing stopped — instead of the contentless “The file structure is invalid or unsupported”.
  • Signed .p7m files now open their contents instead of failing with “Could not extract content from P7M signature”. Signatures that store their document in BER form, in PEM armour, inside an S/MIME wrapper, split across several segments or nested in a second signature are all read now, and the document inside is converted normally. An encrypted .p7m, a detached signature that holds no document, and digest-only files each say which of those they are.
·Cookie and ad consent
  • Visitors in the EEA, the UK and Switzerland now get a cookie notice before any advertising or analytics storage runs, with Accept all, Reject optional and per-category preferences.
  • The choice can be changed or withdrawn at any time from the “Cookie settings” link in the footer, on every page.
  • Google AdSense is installed site-wide, and the ad code requests non-personalised ads until you allow advertising — declining keeps personalisation off instead of blocking the page.
  • The Privacy Policy explains Google AdSense cookies, the opt-out links, and how to turn personalised ads off.
  • The Terms of Service gained an advertising section covering how ads are selected, what is never shared with advertisers, and how to report a misleading ad.
·Conversion fixes
  • EPS files now convert instead of failing with “vector content cannot be rasterized”. Previews stored behind a DOS EPS binary header, behind a corrupt element length, as JPEG thumbnails, and as hex “%%BeginPreview” bitmaps are all decoded. An EPS that truly holds vector artwork only now says so and explains how to convert it, rather than blaming a missing toolkit.
  • macOS .icns icons whose internal element lengths are wrong now convert — the embedded PNG, JPEG or JPEG 2000 payload is located even when the icon's element table is damaged.
  • The admin “Failed conversions” panel no longer shows “NaN failures” with blank Fails, Attempts and Rate cells; those columns now report the real per-pair numbers and totals.
·Conversion fixes
  • macOS .icns icons now convert instead of failing with “contains no decodable image”. Icons stored as raw pixels (ic04/ic05 and the 32-bit variants), icons with 8-bit alpha masks, and icons whose only image is a bare JPEG 2000 codestream are all decoded now.
  • SPSS .sav datasets that contain column headers but no data rows convert successfully — they are valid datasets, so they no longer report “contains no data rows”.
  • Interrupted SPSS .sav uploads now salvage the records that were fully received instead of failing outright. Files truncated before their header is readable still report a clear error.
·Conversion fixes
  • JSON and CSV to M3U playlists now produce real entries. Records without a url column were silently skipped, so the download contained nothing but the #EXTM3U header.
  • Data and document to PDF downloads are far smaller — plain text no longer embeds a 1 MB font, cutting a typical CSV-to-PDF from about 570 KB to under 1 KB.
  • PDF to JFIF now returns a real JPEG instead of PNG bytes with the wrong extension, and an impossible raster target reports a clear error rather than downloading a mislabelled file.
  • Uncompressed BLP2 (Blizzard) textures now convert instead of failing with an unsupported-encoding error.
  • Failed conversions no longer show a generic “tech team notified” line — you now get the specific reason.
·Conversion fixes
  • SPSS .sav uploads that still failed — files with a UTF-8 BOM prefix or extra leading bytes — now parse correctly. Non-SPSS files renamed to .sav (ZIP, Office documents, CSV, SPSS Portable) now say what they actually are instead of failing with a generic error.
·New formats as outputs
  • Recently added formats can now be used as conversion outputs — not just inputs. Examples: MP3→AIFF, MP4→M2TS, ZIP→APK/JAR, CSV→NDJSON, TXT→LOG, PNG→JFIF, and cross-converts between the new audio/video/archive types.
·New formats & homepage
  • Homepage popular pairs expanded to 15 — including MP4→MP3, PDF→PNG, and DOCX→PDF shortcuts below the hero.
  • New image inputs: JFIF and WBMP.
  • New audio inputs: AIFF, AIF, CAF, AC3, and AMR (convert to MP3, WAV, FLAC, and more).
  • New video inputs: M2TS, MTS, VOB, and ASF (convert to MP4, MKV, or extract audio).
  • New archive inputs: APK and JAR (ZIP-based — extract or repackage to ZIP/TAR).
  • New data input: NDJSON (newline-delimited JSON, same as JSONL).
  • New document input: LOG files (convert to PDF, HTML, DOCX, and more).
·Conversion fixes
  • SPSS .sav files now convert reliably to CSV, JSON, and other data targets — including compressed (ZSAV) files and datasets that previously failed with a generic error. A second ReadStat-based reader is used automatically when the primary parser cannot read the file.
·Conversion fixes
  • Parquet files now convert reliably to CSV, TXT, PDF, and other data targets — GZIP, ZSTD, Brotli, and LZ4 compression are supported (previously many real-world Parquet uploads failed).
  • AZW3 and MOBI ebooks convert to Markdown and other text formats without requiring Calibre on the server — a built-in Kindle/KF8 reader extracts text when Calibre is unavailable.
·Conversion fixes
  • DrawIO (.drawio) diagrams now convert to PNG, JPG, and SVG by rendering shapes and connectors — previously PNG export failed unless the file already contained an embedded image.
·Conversion fixes
  • BLP (Blizzard texture) files now convert correctly. Palettized, DXT, and uncompressed BLP2 textures that previously failed with “BLP2 compression 0 not supported” can be converted to BMP, AVIF, PNG, and other image formats.

Found a broken conversion?

Tell us via the contact form — include the input format, target format, and what went wrong. We fix real-world uploads as fast as we can.

Back to the converter, or read about Convertek.

Convertek · www.convertek.tech