I'm looking to learn about the various settings and how each of them
influence the mixes I can create. I've been looking at the code in
*app.py*, especially the similar_api() function.
Taking a look at a snippet from my log, I have a some questions.
Code:
--------------------
PS G:\music-similarity> python .\music-similarity.py -l DEBUG
2022-01-16 14:08:56 D musly.lib set to windows\mingw64\libmusly.dll
2022-01-16 14:08:56 D essentia.highlevel set to False
2022-01-16 14:08:56 D Init DB
2022-01-16 14:08:56 D Start server
2022-01-16 14:08:56 D Init Musly
[...]
2022-01-16 14:08:56 D Musly init done
2022-01-16 14:08:57 D Read jukebox:
G:\music-similarity-data\music-similarity.jukebox
2022-01-16 14:08:57 I Loaded 20568 tracks
2022-01-16 14:08:57 D Ready to process requests
2022-01-16 14:16:09 D Request: {"nogenrematchadj": "15", "genregroupadj":
"7", "shuffle": 1, "maxbpmdiff": "20", "count": "2", "min": 0, "genregroups":
null, "track":
["file:///mnt/PiSounds/flac_mp3/lossless/The%20Beautiful%20South/0898/1992%20-%201-1%20-%20Old%20Red%20Eyes%20Is%20Back.flac"],
"previous":
["file:///mnt/PiSounds/flac_mp3/lossless/The%20Beautiful%20South/0898/1992%20-%201-1%20-%20Old%20Red%20Eyes%20Is%20Back.flac"],
"max": 0, "maxloudnessdiff": "1", "format": "text", "norepart": "25",
"filtergenre": 0, "norepalb": "25", "filterattrib": "1", "filterxmas": "1",
"attribweight": "100", "filterkey": "1", "mpath": "/mnt/PiSounds/flac_mp3"}
2022-01-16 14:16:09 D Genre cfg: {}
2022-01-16 14:16:09 D Essentia(attrib) cfg: {"enabled": true, "highlevel":
false, "bpm": 20, "loudness": 0.1, "filterkey": true, "filterattrib": true,
"weight": 0.0}
2022-01-16 14:16:09 D USABLE ID:9650 Path:lossless/Grease/1978 - 1-19 - Louis
St. Louis - Rock 'n' Roll Party Queen.flac Similarity:0.042277
AdjSim:0.19227686300873756 Meta:{"title": "summer nights", "artist": "grease",
"album": "grease", "albumartist": "grease", "duration": 131, "genres":
["Soundtrack"], "ignore": false, "bpm": 144, "loudness": 0.898295998573, "key":
"FM"}
2022-01-16 14:16:09 D FILTERED(attribs(bpm - 148/93)) ID:15595
Path:lossless/Nabiha/Mind The Gap/2013 - 1-9 - Halo & Handcuffs.flac
Similarity:0.048016 Meta:{"title": "halo and handcuffs", "artist": "nabiha",
"album": "mind the gap", "albumartist": "nabiha", "duration": 227, "genres":
["Pop"], "ignore": false, "bpm": 93, "loudness": 0.937825024128, "key": "Cm"}
2022-01-16 14:16:09 D FILTERED(attribs(loudness -
0.835620343685/0.615452826023)) ID:19639 Path:lossless/Whitesnake/Slip Of The
Tongue/1989 - 1-10 - Sailing Ships.flac Similarity:0.048795 Meta:{"title":
"sailing ships", "artist": "whitesnake", "album": "slip of the tongue",
"albumartist": "whitesnake", "duration": 363, "genres": ["Rock"], "ignore":
false, "bpm": 161, "loudness": 0.615452826023, "key": "GM"}
2022-01-16 14:16:09 D FILTERED(attribs(key - CM (8B) / Gm (6A))) ID:3105
Path:lossy/Dire Straits/Money For Nothing/1988 - 1-11 - Money For Nothing.mp3
Similarity:0.050161 Meta:{"title": "money for nothing", "artist": "sting;dire
straits", "album": "money for nothing", "albumartist": "dire straits",
"duration": 246, "genres": ["Rock:Classic"], "ignore": false, "bpm": 134,
"loudness": 0.829533398151, "key": "Gm"}
[...]
--------------------
1. My music-similarity.db does have high-level analysis data in it.
Because my API server is running on Windows, it appears as if high-level
analysis is turned off: "essentia.highlevel set to False". Because there
is high-level analysis in my database, is this entry accurate? Will it
affect what mixes my system will produce?
2. It appears that tracks are filtered based on high-level attributes.
Is that how I should read the "FILTERED" lines?
- bpm - 148/93. Seed song has BPM of 148, this track is 93. Max
difference setting is 20. 93 is greater than 20 from 148, therefore
filtered based on BPM rule.
- loudness - 0.835620343685/0.615452826023. Similar reasoning, except
loudness setting difference was 0.1.
- key - CM (8B) / Gm (6A) - Similar reasoning. Rule was to find songs
in the same key (CM). What I couldn't find is where the values 8B and
6A were coming from or their purpose, if any.
3. Is there a way to adjust for other attributes like party, electronic,
aggressive, tonal, voice, etc.?
4. There is an instance of a USABLE track in the list. It has these
scores: Similarity:0.042277 AdjSim:0.19227686300873756. These are
relative scores calculated from values in the settings along with values
from music-similarity.db. We see the calculated score for the track
being compared. I couldn't find a similar score for the SEED track. Is
there one? If not, could there be one?
I'd like to be able to see the two scores relative to each other. It
would help me understand how Essentia/Musly/Music-Similarity are working
so that I can best manipulate it.
------------------------------------------------------------------------
prabbit's Profile: http://forums.slimdevices.com/member.php?userid=11142
View this thread: http://forums.slimdevices.com/showthread.php?t=115609
_______________________________________________
plugins mailing list
plugins@lists.slimdevices.com
http://lists.slimdevices.com/mailman/listinfo/plugins
