Losslessly Transcode JPEG to JPEG XL

Walk a directory of JPEG files and convert them to JPEG XL (.jxl) format using lossless encoding. cjxl automatically detects JPEG bitstream and preserves it without re-compression. Reports byte savings; explains how to recover the original JPEG if needed.

When to use

• Archiving JPEG photo libraries to save space (typically 15–25% reduction) without image degradation.
• Preparing a photo collection for long-term storage where compatibility with JPEG is not required.
• Building a smaller-footprint JXL archive that can still yield the original JPEG via djxl --jpeg.

Inputs to gather

• Source directory — folder containing JPEG files (.jpg or .jpeg). Default: current directory.
• Recursive? Ask whether to process subdirectories. Default: no (flat, current folder only).
• Output location — optional. Default: same directory as originals (.jxl files sit alongside .jpg).
• Dry run first? Recommend a preview pass showing the count and byte savings estimate, before committing to the conversion.

Procedure

1. Check tooling

command -v cjxl || echo "cjxl not installed — install with: sudo apt install libjxl-tools"

Abort if missing.

2. Preview

Show the user what will happen:

# Count JPEGs in scope
find <target> -maxdepth <1|999> -type f \
  \( -iname '*.jpg' -o -iname '*.jpeg' \) | tee /tmp/jpegs.txt | wc -l

# Estimate original size
find <target> -maxdepth <1|999> -type f \
  \( -iname '*.jpg' -o -iname '*.jpeg' \) -exec du -ch {} + | tail -1

Print:

• Number of JPEG files found.
• Total original size.
• Estimated output size (advise 20% reduction as a rough baseline).
• Confirm before proceeding.

3. Transcode

Use -d 0 (lossless) and a reasonable effort (suggest 6–7 for balance):

Folder (non-recursive):

find <target> -maxdepth 1 -type f \
  \( -iname '*.jpg' -o -iname '*.jpeg' \) \
  -exec cjxl -d 0 -e 7 {} {.}.jxl \;

Recursive:

find <target> -maxdepth 999 -type f \
  \( -iname '*.jpg' -o -iname '*.jpeg' \) \
  -exec cjxl -d 0 -e 7 {} {.}.jxl \;

(The -d 0 flag ensures lossless encoding. Effort 7 balances speed and compression. Effort 9 is slower but slightly smaller; only recommend if the user has time.)

4. Verify and report

After transcoding:

# Count JXL files created
find <target> -maxdepth <1|999> -type f -iname '*.jxl' | wc -l

# Total size of JXL files
find <target> -maxdepth <1|999> -type f -iname '*.jxl' -exec du -ch {} + | tail -1

# Bytes saved (original size - new size)
du -ch <target> | tail -1

Print:

• Files successfully transcoded.
• Original total size vs. new total size.
• Bytes saved and percentage reduction.
• Any failures (transodes that produced errors).

5. Explain recovery path

Inform the user:

If you ever need to recover the original JPEG bitstream byte-exact (for legal/archival purposes), run:

djxl --jpeg archive.jxl recovered.jpg

This works because cjxl detected and preserved the original JPEG structure.

Output / side effects

• .jxl files created in the source directory (or user-specified output folder).
• Original .jpg files remain untouched.
• No backup is created; if the originals must be preserved, the user should copy them first.

Notes

• Lossless JPEG preservation: cjxl auto-detects JPEG bitstreams and preserves them exactly. This is why djxl --jpeg can recover the original — it's not a re-encoded approximation.
• Effort 7 vs. 9: Effort 7 compresses in seconds per image; effort 9 can take minutes. For batch workflows, 7 is recommended. Use 9 only if compression size is the overriding concern.
• Size savings: Typical JPEG → JXL lossless reduction is 15–25%, depending on the original JPEG quality. Do not oversell — some JPEGs may compress only 5–10%.
• Selective transcoding: If you want to skip already-small files or transcode only above a size threshold, filter the find output (e.g., -size +100k for files larger than 100 KB).
• Do not delete the original JPEG files automatically. Let the user confirm removal after verifying the JXL output.
• Depends: apt install libjxl-tools.