cleanorlabs/cleanor-storage-lab
pretty_name: "Cleanor Storage Lab" license: cc-by-4.0 language: - en tags: - image-compression - avif - webp - jpeg-xl - heic - cloud-storage - benchmark - open-data size_categories: - n<1K configs: - config_name: compression-benchmark data_files: compression-benchmark.csv - config_name: heic-tax data_files: heic-tax-benchmark.csv - config_name: nextgen-formats data_files: nextgen-formats-benchmark.csv - config_name: cloud-price-index… See the full description on the dataset page: https://huggingface.co/datasets/cleanorlabs/cleanor-storage-lab.
<!-- Ready-to-use Hugging Face dataset card.
How to publish:
- Create a dataset at https://huggingface.co/new-dataset (owner: cleanor, name: storage-lab)
- Upload the CSVs from this repo's
data/folder. - Paste EVERYTHING below (including the --- YAML block) as the dataset's README.md. Keep this file in sync with the GitHub README when the data updates. --> --- pretty_name: "Cleanor Storage Lab" license: cc-by-4.0 language:
- en tags:
- image-compression
- avif
- webp
- jpeg-xl
- heic
- cloud-storage
- benchmark
- open-data size_categories:
- n<1K configs:
- configname: compression-benchmark datafiles: compression-benchmark.csv
- configname: heic-tax datafiles: heic-tax-benchmark.csv
- configname: nextgen-formats datafiles: nextgen-formats-benchmark.csv
- configname: cloud-price-index datafiles: cloud-price-index.csv
- configname: photo-storage-capacity datafiles: photo-storage-capacity.csv
- configname: video-codec-benchmark datafiles: video-codec-benchmark.csv
- configname: video-codec-summary datafiles: video-codec-summary.csv
- configname: app-size-index datafiles: app-size-index.csv
- configname: storage-upgrade-tax datafiles: storage-upgrade-tax.csv
- configname: gif-format-benchmark datafiles: gif-format-benchmark.csv
- configname: lossless-recompression datafiles: lossless-recompression.csv
- configname: font-format-weight datafiles: font-format-weight.csv
- configname: svg-minify datafiles: svg-minify.csv
- configname: pdf-compression datafiles: pdf-compression.csv ---
Cleanor Storage Lab
Reproducible benchmarks and open datasets on image compression (JPEG / WebP / AVIF / JPEG XL), the HEIC-to-JPG size tax, cloud storage price per GB, and phone storage capacity.
Every number is generated by a script you can re-run yourself. Image benchmarks run on the Kodak and public high-resolution phone-photo corpora, encoded with libvips/sharp, libjxl and libheif, and scored with SSIM + PSNR against the lossless master.
Highlights (quick answers)
- AVIF vs JPEG: at identical quality (SSIM 0.95), AVIF is ~41% smaller than JPEG; WebP ~13% smaller.
- HEIC to JPG at matched quality makes a photo corpus ~2.5x bigger (to PNG, ~5.5x) — the "HEIC tax".
- Xvid/DivX vs modern video codecs: Xvid needs ~4x the bitrate of H.264 and ~8.9x AV1 at the same quality.
- GIF vs MP4: the same clip is 44-91x smaller as an MP4 than a GIF.
- App sizes: the median popular iOS app is ~400 MB to download; TikTok is 936 MB.
- Storage upgrade tax: makers charge ~$0.78/GB to add storage, ~13x the ~$0.06/GB flash cost.
- Lossless recompression: oxipng shrinks a PNG ~10%, jpegtran a JPEG ~6%, bit-identical.
- Fonts: a WOFF2 is ~40% of the TTF; subset to Latin, ~11%.
- PDF: Ghostscript /ebook cuts ~70% (lossy, text kept); qpdf ~13% lossless.
Files
Sources & docs
- Cite: DOI 10.5281/zenodo.21217372 (concept DOI, all versions)
- Code + harness: https://github.com/cleanor-app/cleanor-storage-lab
- Open datasets catalog: https://cleanor.app/research/open-datasets
- Human-readable studies with charts: https://cleanor.app/research
License
Data: CC BY 4.0 (attribution required). Code in the source repo: MIT.
Citation
Cleanor Labs. Storage Lab: Open Image-Compression & Cloud-Storage Datasets. 2026. https://cleanor.app/research
