From 1695953705d08924a76b6032cac1502d8a6779f2 Mon Sep 17 00:00:00 2001
From: Broque Thomas <26755000+Nezreka@users.noreply.github.com>
Date: Mon, 20 Apr 2026 12:46:01 -0700
Subject: [PATCH] Fix AcoustID high-confidence skip letting wrong files through
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

The high-confidence fingerprint skip (≥0.95) assumed title mismatches
were language/script differences and bypassed verification. But a high
fingerprint score just means AcoustID identified the audio confidently —
not that it matches the requested track. Now requires partial title
(≥0.55) or artist (≥0.60) similarity before skipping, so completely
wrong files (e.g. different song/artist from same remix producer) are
correctly rejected.
---
 core/acoustid_verification.py | 10 ++++++----
 1 file changed, 6 insertions(+), 4 deletions(-)

diff --git a/core/acoustid_verification.py b/core/acoustid_verification.py
index f3f9d270..5bfb6f92 100644
--- a/core/acoustid_verification.py
+++ b/core/acoustid_verification.py
@@ -330,11 +330,13 @@ class AcoustIDVerification:
                     logger.info(f"AcoustID verification PASSED (scan match) - {msg}")
                     return VerificationResult.PASS, msg
 
-            # No match found — but if fingerprint score is very high (≥0.95),
-            # the audio IS correct and the title mismatch is likely a language
-            # difference (e.g. Japanese kanji vs English title for the same song).
+            # No match found — but if fingerprint score is very high (≥0.95)
+            # AND there's partial similarity in title or artist, the mismatch is
+            # likely a language/script difference (e.g. Japanese kanji vs English).
             # Skip rather than quarantine a correct file.
-            if best_score >= 0.95:
+            # But if both title AND artist similarity are very low, the download
+            # source gave us a completely wrong file — fail it.
+            if best_score >= 0.95 and (title_sim >= 0.55 or artist_sim >= ARTIST_MATCH_THRESHOLD):
                 top = recordings[0]
                 msg = (
                     f"Title/artist mismatch but fingerprint confidence very high ({best_score:.2f}): "