Skip to content
FlipMyFile

Changelog

What has shipped, in the order it shipped — including the mistakes we found in our own measurements and what the corrected numbers turned out to be.

  1. Format reference, tool documentation and structured data

    • AddedA file format reference at /formats covering JPG, PNG, WebP, AVIF, HEIC, GIF, MP4, MOV, WebM and MP3 — MIME types, compression, transparency, metadata and browser support, each linked to the tools that read or write it.
    • AddedEvery tool page now publishes its input and output formats, size, resolution and duration ceilings, processing engine and retention window as a table read from the same configuration the server enforces.
    • AddedOutput quality, common use cases and troubleshooting sections written individually for all 25 tools, plus an About page, this changelog, and a security page describing the full life of an uploaded file.
    • ChangedStructured data now declares one Organization and one WebSite for the whole site, with each tool page describing itself as a WebApplication that references them, instead of 25 unconnected fragments.
    • FixedThe sitemap claimed every page had been updated on each deploy because it used the build date. It now uses real review dates, and result-sharing links (?jobs=…) are excluded from crawling so they cannot create an unlimited set of URLs.
    • FixedThe privacy page said traffic passes through Cloudflare as a CDN. Cloudflare provides DNS for the domain, but requests reach our server directly, so the description has been corrected.

  2. Image quality: three measurement bugs found and fixed

    • FixedJPEG output used 4:2:0 chroma subsampling for every image, which smeared saturated colour and coloured text. Above quality 70 the full colour resolution is now kept: on the affected test image, structural similarity rose from 0.9733 to 0.9999 and peak signal-to-noise ratio by 18.6 dB, at a cost of 4 KB.
    • FixedPNG palette compression banded gradients (similarity 0.9892, average colour error 3.04). The compressor now measures the palette version and falls back to lossless below 0.995.
    • FixedCompression quality was being measured on a 384-pixel copy to save time. That reported 0.9913 where the true figure at full size was 0.9783 — downscaling averages neighbouring pixels and hides exactly the edge damage the measurement exists to catch. Measurement now runs at full resolution up to 2 megapixels.
    • ChangedQuality sliders were replaced by three modes that map to similarity thresholds (0.998 / 0.99 / 0.98). The worker searches for the lowest quality setting that still clears the threshold for your particular image, because the same number means different things on a smooth portrait and on foliage.
    • ChangedCompression tools never return a larger file: if no setting beats the original, the original is returned untouched.

  3. Video and audio tools

    • AddedEleven video and audio tools: MOV to MP4, WEBM to MP4, compress, resize, rotate, mute, change speed, reverse, trim, MP4 to MP3 and MP4 to GIF.
    • FixedConcurrency slots were counted in the database before the job was written, so simultaneous uploads could all pass a limit of three. Slots are now reserved in Redis in a single operation before the count is read, with orphaned reservations released after 20 minutes.
    • FixedA shared ceiling of two concurrent jobs rejected five images out of nine when a folder was dropped in. Image limits were raised to match how batch conversion is actually used, and the client now queues uploads three at a time and retries automatically instead of showing an error.
    • ChangedReverse Video is limited to 15 seconds. The FFmpeg reverse filter cannot stream — it holds every decoded frame in memory — and 15 seconds of 720p already exceeds a gigabyte.

  4. Launch

    • AddedFlipMyFile went live with image conversion, compression, resizing and cropping, each tool on its own page with its own settings and limits.
    • AddedHelp centre, contact form and blog.
    • ChangedUsage figures on tool pages are read from real job records and hide themselves below 25 jobs, rather than displaying invented review counts and ratings.

Found something that behaves differently from what is described here? Tell us.