Lightroom Classic: Optimize backup compression for AI data (.lrcat-data)

Feature Request: Optimize Lightroom Classic Backup Compression for AI Data

I am a professional photographer in Brazil, and I would like to report a significant inefficiency in Lightroom Classic's catalog backup process.

After experiencing extremely slow backups, I conducted a controlled performance test to identify the bottleneck.

Test environment

• Software: Adobe Lightroom Classic 15.6

• Operating system: Windows 11

• RAM: 32 GB

• Primary storage: Kingston NV3 2 TB PCIe 4.0 NVMe SSD

• Main catalog (.lrcat): Approximately 307 MB

• AI data directory (.lrcat-data): 55.6 GB

• Final backup ZIP: 54.5 GB

Measured backup performance

Using a stopwatch, I recorded each stage of the built-in backup:

• Initial preparation: 2 minutes 16 seconds

• Additional processing: 2 minutes 42 seconds

• Catalog optimization: 2 minutes 44 seconds

• ZIP compression: 25 minutes 54 seconds

• Total backup time: 33 minutes 40 seconds

Compression alone accounted for approximately 77% of the entire backup operation.

During compression, overall CPU utilization was generally around 19–31%, RAM usage remained approximately 14 GB, and NVMe activity was frequently low.

Technical findings

After inspecting the generated ZIP archive, I confirmed that most of its contents consisted of .blob files from the .lrcat-data directory.

The inspected files showed compression ratios of approximately 2–4%, meaning that the compression process achieved very little storage reduction despite consuming almost 26 minutes.

The main catalog database was much smaller and showed significantly better compression.

These results suggest that the backup process spends substantial time compressing AI-related data that is already difficult to compress.

Proposed improvement

I suggest implementing an adaptive backup compression mechanism.

Specifically:

1. Detect files or data blocks with low compressibility.

2. Use ZIP's Store method (without compression) for nearly incompressible .blob files.

3. Continue compressing the main .lrcat database and other compressible data.

4. Preserve existing catalog integrity checks, backup compatibility, and restoration reliability.

5. Optionally provide a user-selectable Fast Backup / No Compression mode.

The goal is not to exclude AI data from backups, but to avoid wasting processing time compressing data that offers minimal storage savings.

Why this matters

Professional photographers increasingly depend on AI Denoise, intelligent masks, generative removal, and other AI-assisted editing tools.

As .lrcat-data grows, catalog backups can become increasingly time-consuming.

In my case, the complete backup required more than 33 minutes even when performed entirely on an NVMe SSD.

This creates a significant workflow bottleneck for professionals who need frequent, reliable catalog backups.

A more efficient compression strategy could substantially improve backup performance without compromising data integrity.

A similar issue has also been documented by other Lightroom Classic users in the Adobe Community:

https://community.adobe.com/topic/show?tid=1634033&fid=675

I hope these independently measured results help the Lightroom Classic engineering team investigate and improve this behavior.

Thank you for considering this request.