Release notes

What changed in each version of Wavebase.

v0.1.3 — July 2026

The AI review release.

AI review assistant. After each scan, a local language model re-examines every file flagged for review. When it disagrees with the first classification it proposes a fix — an amber wand next to the file applies it with one click, and turns into a revert arrow if you change your mind. A green check means the AI double-checked the row and agrees. Everything runs on your machine: no cloud, no account.
Instant results table. The review table now appears the moment audio analysis finishes; AI suggestions stream in the background with their own progress bar while you already work.
Scan a folder of folders. If your source folder contains several audio subfolders, Wavebase asks whether you want one combined tab or one tab per subfolder — each collection stays in its own workspace, including loose top-level files.
Bigger free trial. The trial now lets you export metadata for up to 150 assets (previously 50).
Resilient AI model install. The assistant's model (~1.9 GB) downloads automatically in the background — also after updating from an older version. If a firewall or VPN blocks the download, Wavebase stays fully usable and offers a one-click install later, with a live progress bar.
Faster, steadier scans. Audio inference is cached (one pass per file instead of up to four), and a long list of stability fixes: safer database concurrency, serialized AI processing, and per-file error reporting.

Much better with music. Stingers, cues and tracks are now told apart by actual duration; BPM and key are detected based on what the audio really is (not just its label), with octave-corrected tempo (no more 60 vs 120 confusion).
Fewer false positives from filenames. Words hiding inside other words ("truck" inside "bowtruckle") no longer force a wrong category.
Scans that cannot hang. Per-file timeouts, unreadable-file detection with a skipped-files report, and steadier memory use on long batches.
Search bar to filter the review table by file name.

Naming conventions. Load a Markdown file with your studio's abbreviations and vocabulary; Wavebase reads your library the way your team named it.
BPM and musical key for music files, right alongside the rest of the metadata.
Save your session and reopen exactly where you left off.

First public release: scan your library, review AI-generated metadata, embed it into your WAV files, export CSV.

Using Wavebase and missing something? Tell us what you think — the roadmap is built from what real libraries need.