Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Sign up for free
Menu
Search
Features
All features
Private URLs
Password Protection
Custom URLS
Scheduled publishing
Remove Branding
Restrict embedding
Deck Collections
Notes
Features
All features
Private URLs
Password Protection
Custom URLS
Scheduled publishing
Remove Branding
Restrict embedding
Deck Collections
Notes
Explore
Featured decks
Featured speakers
Programming
Technology
Storyboards
Explore
Featured decks
Featured speakers
Programming
Technology
Storyboards
Pricing
Search
Sign in
Sign up for free
Rubyで音を視る
Search
Sponsored
·
Ship Features Fearlessly
Turn features on and off without deploys. Used by thousands of Ruby developers.
→
ydah
June 06, 2026
Technology
730
1
Share
Embed
Copy iframe code
Copy JS code
Copy link
Start on current slide
Rubyで音を視る
松江Ruby会議12「Rubyで音を視る」の発表スライド
https://matsue.rubyist.net/matrk12/
#matrk12
ydah
June 06, 2026
More Decks by ydah
See All by ydah
わからない話を追いかけたら、プログラミング言語を作る側にいた
ydah
3
1k
Liberating Ruby's Parser from Lexer Hacks
ydah
2
4.2k
ルールルルルルRubyの中身の予備知識 ── RubyKaigiの前に予習しなイカ?
ydah
1
1.1k
計算機科学をRubyと歩む 〜DFA型正規表現エンジンをつくる~
ydah
3
980
進捗
ydah
2
360
Joy with 3D Graphics Using Ruby
ydah
1
170
構文解析器入門
ydah
7
3.7k
Regional.rb and the Kyoto City
ydah
0
97
LRパーサーはいいぞ
ydah
7
2.1k
Other Decks in Technology
See All in Technology
継続的なAIコスト最適化
bgpat
1
210
1万名の社員が使う認証基盤で どう信頼性を担保するか?
kairim0
0
210
AIに攻撃される前に、AIに攻撃させる
tsuchikazu
0
160
しろおびから始める!ログ活用の第一歩
shumei_ito
0
510
自分で立ててみるLLMサービス
y_sera15
0
140
初対面のメンバーがチームを組み、Claude Codeでゼロからサービスをリリース
mkdev10
0
100
自主式軟體工廠
philipz
0
240
生成AIで高い生産性を求められて、 悩み、乗り越えた話
mot_techtalk
2
130
今こそ知りたいAmplifyGen2
mkdev10
2
160
私の推しは「聞いてから進む」AIです -AI-DLCに一人でアプリを作らせた話
yama3133
1
180
いちAWSエンジニアのAI活用を振り返る #devio2026 / devio osaka 2026 kawahara
masahirokawahara
2
410
Futexes the good, the bad, the ugly
ennael
PRO
0
130
Featured
See All Featured
Kristin Tynski - Automating Marketing Tasks With AI
techseoconnect
PRO
0
540
Designing Dashboards & Data Visualisations in Web Apps
destraynor
232
55k
個人開発の失敗を避けるイケてる考え方 / tips for indie hackers
panda_program
123
22k
Jamie Indigo - Trashchat’s Guide to Black Boxes: Technical SEO Tactics for LLMs
techseoconnect
PRO
0
690
Have SEOs Ruined the Internet? - User Awareness of SEO in 2025
akashhashmi
0
510
sira's awesome portfolio website redesign presentation
elsirapls
0
440
Documentation Writing (for coders)
carmenintech
77
5.6k
Hiding What from Whom? A Critical Review of the History of Programming languages for Music
tomoyanonymous
3
1.3k
What Being in a Rock Band Can Teach Us About Real World SEO
427marketing
0
1.1k
ラッコキーワード サービス紹介資料
rakko
1
5.2M
Building a Modern Day E-commerce SEO Strategy
aleyda
45
9.2k
GraphQLの誤解/rethinking-graphql
sonatard
75
12k
Transcript
2026/06/06 Ruby 12 Ruby seeing sound with ruby Yudai Takada,
ydah
None
None
None
None
None
Otsu City Traditional Performing Arts Hall Saturday, July 18, 2026
None
None
None
None
None
None
2026/06/06 Ruby 12 Ruby seeing sound with ruby Yudai Takada,
ydah
None
None
vizcore https://github.com/ydah/vizcore
Vizcore visualize( ) × core( ) Audio-reactive visuals Ruby DSL
Ruby scene renderer
2D/3D OpenGL
None
None
None
None
1
3
PCM samples Mic / File samples Array<Float> [ 0.0012, 0.0031,
-0.0028, . . . ]
None
Sample Window tick window 1024 60fps tick sample 圧
Samples RMS amplitude Sample RMS 夏 圧 0 [0.5, -0.5,
0.5, -0.5] (0.5 + -0.5 + 0.5 + -0.5) / 4 = 0 RMS = 1 N N ∑ i=1 x2 i
Samples FFT magnitudes / peak_frequency FFT 圧 WN = e−2πi/N
Ek = N/2−1 ∑ m=0 x2m Wkm N/2 Ok = N/2−1 ∑ m=0 x2m+1 Wkm N/2 Xk = Ek + Wk N Ok Xk+N/2 = Ek − Wk N Ok (k = 0,1,…, N 2 − 1)
magnitudes BandSplitter sub / low / mid / high FFT
Vizcore sub, low, mid, high sub: 20Hz . . 60Hz low: 60Hz . . 250Hz mid: 250Hz . . 4000Hz high: 4000Hz . . 20000Hz
Samples frame energy BeatDetector beat FFT beat beat frame 圧
E = 1 N N ∑ i=1 x2 i
beat beat_pulse beat boolean frame true 1 frame beat 圧
1.0 beat frame 0.86 長
beat history BPMEstimator bpm / beat_phase / bar_phase BPM beat
frame lag BPM beat beat beat = beat_pulse = bpm = beat_phase = bar_phase = bpm = 60.0 * frame_rate / best_lag
bands + onset drums con fi dence 圧 band onset
kick = sublow_strength * onset # sub/low ͷڧ͞ × onset snare = mid_strength * onset # mid ͷڧ͞ × onset hihat = high_strength * onset # high ͷڧ͞ × onset
magnitudes / spectrum spectral features FFT Spectral features FFT 圧
amplitude bands spectral features centroid: / rollo ff : / fl atness: / fl ux: /
raw features smoother audio features frame を EMA EMAt =
αxt + (1 − α)EMAt−1 amplitude: @sm.smooth(:amplitude, normalized[:amplitude]) bands: @sm.smooth_hash(normalized[:bands], namespace: :bands) fft: @sm.smooth_array(normalized[:fft], namespace: :fft)
amplitude bass / low high / treble beat beat_pulse bpm
/ phase spectral_* / / / / / を 長
None
audio features MappingResolver layer params map beat_pulse, to: :radius, gain:
160.0, min: 56, max: 164, attack: 1.0, release: 0.2
beat beat_pulse map beat_pulse, to: :radius, gain: 100, min: 50,
max: 150
beat map beat_pulse, to: :radius, gain: 100, min: 50, max:
150 100
beat map beat_pulse, to: :radius, gain: 100, min: 50, max:
150 50 150
beat map beat_pulse, to: :radius, gain: 100, min: 50, max:
150 radius
layer params audio_frame audio scene layers audio frame { “audio”:
{ "amplitude": 0.42, "beat": true, "beat_pulse": 1.0, "bands": { "low": 0.8 } }, "scene": { "name": "main", "layers": [ { "name": "rings", "params": { "radius": 140 } …
audio_frame WebSocket frontend WebGL renderer Frontend audio_frame Ruby Runtime Browser
Renderer (WebGL Runtime) Web Socket
6/4
BGM
None
None
50 6 4
2026/06/06 Ruby 12 Ruby seeing sound with ruby Yudai Takada,
ydah
圧 母 /
母
母 母 F1 / F2 spectral centroid low / mid
/ high ratio
F1/F2 母 F 圧 圧 : : : / :
:
母 0.0 1.0 1 F1 F2 母
None
/ / 圧 圧 burst fl atness ZCR nasal high
ratio
burst amplitude_delta = [ current[:amplitude] - previous[:amplitude], # લϑϨʔϜ͔ΒͷԻྔ૿Ճ 0.0
].max burst_strength = [ current[:onset], # Իͷ্ཱ͕ͪΓ current[:spectral_flux], # εϖΫτϧมԽྔ current.dig(:onsets, :mid), # தҬͷ্ཱ͕ͪΓ current.dig(:onsets, :high), # ߴҬͷ্ཱ͕ͪΓ amplitude_delta, # Իྔͷ૿Ճྔ recent_onset_max # ۙ૭ͷ࠷େ্ཱ͕ͪΓ ].max
spectral fl atness s_score = weighted_average([ high_ratio, # ߴҬͷڧ͞ flatness,
# ϊΠζͬΆ͞ zcr, # ܗͷࡉ͔͍༳Ε centroid # Իͷ໌Δ͞ ]) n_score = weighted_average([ nasal_score, # ඓԻͬΆ͞ 1.0 - high_ratio, # ߴҬ͕গͳ͍ low_mid_ratio, # ʙதҬͷڧ͞ 1.0 - zcr # ༳Ε͕গͳ͍ ])
attack feature attack = { strength: attack[:attack_strength], # ग़ͩ͠ͷڧ͞ centroid:
attack[:centroid_norm], # ग़ͩ͠ͷ໌Δ͞ high: attack[:high_ratio], # ߴҬͷڧ͞ low: attack[:low_ratio], # Ҭͷڧ͞ flatness: attack[:spectral_flatness],# ϊΠζͬΆ͞ zcr: attack[:zero_crossing_rate], # ࡉ͔͍༳Ε f1 : attack[:f1_frequency], # F1 ͬΆ͍ࢁ f2 : attack[:f2_frequency], # F2 ͬΆ͍ࢁ nasal: attack_nasal_score # ඓԻͬΆ͞ }
母 × kana_score = [ vowel_score * 0.55, # ԻΒ͠͞
consonant_score * 0.35, # ࢠԻΒ͠͞ burst_bonus * 0.10 # ग़ͩ͠ͷิਖ਼ ].sum candidate = { text: kana_table[consonant][vowel], # ͔ͳจࣈ vowel: vowel, # ਪఆԻ consonant: consonant, # ਪఆࢠԻ conf i dence: kana_score # ީิͷࣗ৴ }
None
vizcore Ruby ォ Ruby
$whoami ォ ydah ( ) SmartHR Ruby 西 Ruby Ruby
( 西 ) 西 圧 OSS 2