Upload 2655 files

.gitattributes

@@ -58,3 +58,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 # Video files - compressed
 *.mp4 filter=lfs diff=lfs merge=lfs -text
 *.webm filter=lfs diff=lfs merge=lfs -text
+Data/results.numbers filter=lfs diff=lfs merge=lfs -text

Data/audios/S01-mapto5639-S01_HOLDOUT_01.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0edec976b9efd7becabddaf30fc4478f0133bc009d3f7aab0d1b4971761eb1eb
+size 602604

Data/audios/S01-mapto5639-S01_TRAIN_01.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a55e3ba341669522fabde457a29727dffd1bc108abeb7a8f6ad01525a014fe32
+size 1139884

Data/audios/S01-mapto5639-S01_TRAIN_02.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:443cfc69a8e394a772f2fd6da4d1c35ef35dcce1bda6a859f29877df2a6933c9
+size 1277164

Data/audios/S02-mapto7729-S02_HOLDOUT_01.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c74ad80b21020fef2cd47b45813cfe5381c1e0fc812807f834b4cb007232c1d2
+size 563564

Data/audios/S02-mapto7729-S02_TRAIN_01.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c0eedec7bc81dc5259fab3d147844bd480671004d73c33daef27ebb3a761e451
+size 988204

Data/audios/S02-mapto7729-S02_TRAIN_02.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b7b810ccbbd4fe4ae45be3453af96c4a9024147513e1c36b58464c5a300fff75
+size 1436204

Data/audios/S03-mapto1580-S03_HOLDOUT_01.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:34b9a0a959281c4b14813b97d4b1b3e346658b782a2cc65488ba4cada57e6da3
+size 483884

Data/audios/S03-mapto1580-S03_TRAIN_01.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:90139a5f5ad2ffbcd4efc1f4ee61c1b0eb6b0c2cad83f6c2a61db6d59fa906ad
+size 910124

Data/audios/S03-mapto1580-S03_TRAIN_02.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:75c22af310a155a83b3ea4287a26ef5f0feb8550ffa1c322798e5c98bd602a47
+size 1040684

Data/audios/S04-mapto4077-S04_HOLDOUT_01.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:169b608f23370d11929dbf72abea6a23394b2e1a4191890eb19afc2acd12f390
+size 658924

Data/audios/S04-mapto4077-S04_TRAIN_01.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7d0a26aec753fa2086ea7bbe1f0d5e61a4a6c38aa1f8117cd8b7ce5d36a38d81
+size 897324

Data/audios/S04-mapto4077-S04_TRAIN_02.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:93387e3f60d507bdfa09cd2b2c3b1489e4a941d811336a86946d1ebe063bb4c4
+size 1087724

Data/audios/S05-mapto3575-S05_HOLDOUT_01.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:829a64d28f90fe18c0b2a118b0b134f62024ec7021b9f5e1b6dd0df58bb62f79
+size 631404

Data/audios/S05-mapto3575-S05_TRAIN_01.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2f369265fde3b231d51a349e3d614c985378da836353e817bd4df1e27d37345b
+size 962924

Data/audios/S05-mapto3575-S05_TRAIN_02.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0edee2c9f04a96fa402aa5a804966e491974628826e57c89b618ea53f6a658d8
+size 1156524

Data/audios/S06-mapto1995-S06_HOLDOUT_01.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1cf4af4f066941b7ce2d21cfbb17e54510c67bf9ff5efb37d70462901fa140e4
+size 653804

Data/audios/S06-mapto1995-S06_TRAIN_01.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8fe08ac8848af48b24ff21961062be5f55b477906dc8cf5b307a2881943ad90e
+size 820524

Data/audios/S06-mapto1995-S06_TRAIN_02.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dfe71432fbfe09269ce3e3e858d14ec847f3f5484d278131c6e55aa9402b86d7
+size 1239724

Data/audios/S07-mapto2300-S07_HOLDOUT_01.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc88e8d5ebbe7f3718ce1b548a7197b28f12a6bf3a4fb78c925c842a91f1a8ce
+size 550444

Data/audios/S07-mapto2300-S07_TRAIN_01.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1b67d779c1479c160a8fa97b11ff96b31e2688d4a862263c218c2541230ab001
+size 865324

Data/audios/S07-mapto2300-S07_TRAIN_02.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:223e4c57d68a399d84d5d07cd06de7e29c9c19b3b3ef3b57b68e41c7291a34c6
+size 1020524

Data/audios/S08-mapto1221-S08_HOLDOUT_01.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ac1bd07401a7ff410a6cb88e949df1212bc28beb49dea070f2c72f3eb57401b2
+size 494124

Data/audios/S08-mapto1221-S08_TRAIN_01.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8fbc3d75ddbce76c579b221ee9178986e322154667b7fbde2b105aeffab333d3
+size 961324

Data/audios/S08-mapto1221-S08_TRAIN_02.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7bf21e838be06e0b83e14b718c413ba57b94d5e1f7f51fdecf9c6ae1990f6e55
+size 1172524

Data/audios/S09-mapto8463-S09_HOLDOUT_01.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:22ded8a374578b42e732610a5918fd638c77d5e55151d21bfcd02bc1dd39c47c
+size 670124

Data/audios/S09-mapto8463-S09_TRAIN_01.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:49315b72992f3dbae428a32617efdfff9e2b86413306c016405d8869dc31997d
+size 835564

Data/audios/S09-mapto8463-S09_TRAIN_02.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2a7570c91e99219adfb37d7b3a06bccf6634b728a5f90656b5567cd29fecb023
+size 1026924

Data/audios/S10-mapto5105-S10_HOLDOUT_01.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:45f8a5cac676481b090f43b8d3ccc86a7be5d6ec849788f2c8f9085fe6a0d1c5
+size 615724

Data/audios/S10-mapto5105-S10_TRAIN_01.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a1f185ea4b30e32accbfee96c412c905554bf3f0e35c57597b0989cd56c031b5
+size 1132844

Data/audios/S10-mapto5105-S10_TRAIN_02.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7b477e85c7eda75eedadd4655c9aa2982a66dd686d6313c8ca747e04fd876726
+size 1077804

Data/results.numbers

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:022bbc1661775f542604bf50fd86bd56fe8fdf8a39587a3f9245b2a7036d1e15
+size 183209

Data/search_youtube.py

@@ -0,0 +1,722 @@
+#!/usr/bin/env python3
+"""
+Find recent YouTube video URLs with yt-dlp, no API key required.
+
+Requirements:
+    pip install -U yt-dlp
+
+Usage:
+    python find_recent_youtube_urls.py
+
+What it does:
+- Searches YouTube using yt-dlp's ytsearchN:QUERY syntax
+- Applies your query patterns (after:2025-04-01, minus repost terms, etc.)
+- Collects video URL, title, uploader, upload date, duration, view count
+- Writes results to:
+    - videos.csv
+    - videos.json
+"""
+
+from __future__ import annotations
+
+import csv
+import json
+import re
+import sys
+from dataclasses import dataclass, asdict
+from datetime import datetime, date
+from pathlib import Path
+from typing import Iterable, List, Dict, Any, Optional, Set
+
+import yt_dlp
+
+
+# ---------------------------
+# Configuration
+# ---------------------------
+
+CUTOFF_DATE = date(2025, 3, 31)
+AFTER_TOKEN = "after:2025-04-01"
+
+MAX_DURATION_SECONDS = 180
+
+# How many results to ask YouTube for per query
+MAX_RESULTS_PER_QUERY = 30
+
+# Output files
+CSV_OUTPUT = Path("videos.csv")
+JSON_OUTPUT = Path("videos.json")
+
+# Set to True to print every result as it is found
+VERBOSE = True
+
+# Add your real search targets here.
+# These are the concrete things you care about.
+TARGETS: Dict[str, List[str]] = {
+    "person_name": [
+        "Taylor Swift",
+        "Ellie Goulding",
+        "Dua Lipa",
+        "Adele",
+        "Beyoncé",
+        "Billie Eilish",
+        "Drake",
+        "Bad Bunny",
+        "Elon Musk",
+        "Sam Altman",
+        "Jensen Huang",
+        "Sundar Pichai",
+        "Tim Cook",
+        "Mark Zuckerberg",
+        "Keir Starmer",
+        "Donald Trump",
+        "Joe Biden",
+        "Emmanuel Macron",
+        "Volodymyr Zelenskyy",
+        "Cristiano Ronaldo",
+        "Lionel Messi",
+        "Kylian Mbappé",
+        "Erling Haaland",
+        "Novak Djokovic",
+        "Lewis Hamilton",
+    ],
+    "event_name": [
+        "WWDC 2025",
+        "Google I O 2025",
+        "CES 2025",
+        "Paris Fashion Week",
+        "Milan Fashion Week",
+        "London Marathon 2025",
+        "New York Marathon 2025",
+        "Champions League",
+        "Premier League",
+        "FA Cup",
+        "Super Bowl",
+        "NBA Finals",
+        "Oscars 2025",
+        "Grammy Awards 2025",
+        "Cannes Film Festival 2025",
+        "SXSW 2025",
+        "COP30",
+        "UN General Assembly",
+        "G20 Summit",
+        "NATO Summit",
+        "Comic Con 2025",
+        "Gamescom 2025",
+        "Tokyo Game Show 2025",
+        "E3 2025",
+        "Summer Olympics 2025",
+    ],
+    "company_name": [
+        "OpenAI",
+        "Tesla",
+        "NVIDIA",
+        "Apple",
+        "Google",
+        "Alphabet",
+        "Microsoft",
+        "Meta",
+        "Amazon",
+        "SpaceX",
+        "xAI",
+        "Anthropic",
+        "Adobe",
+        "Intel",
+        "AMD",
+        "Samsung",
+        "Sony",
+        "Disney",
+        "Netflix",
+        "TikTok",
+        "ByteDance",
+        "Uber",
+        "Airbnb",
+        "Nike",
+        "Adidas",
+    ],
+    "product_model": [
+        "iPhone 17",
+        "iPhone 16 Pro",
+        "Samsung Galaxy S25",
+        "Google Pixel 10",
+        "Tesla Model Y",
+        "Tesla Model 3",
+        "Tesla Cybertruck",
+        "Canon R5 Mark II",
+        "Sony A7S III",
+        "Sony A7 IV",
+        "DJI Osmo Pocket 3",
+        "DJI Mini 4 Pro",
+        "MacBook Pro M4",
+        "iPad Pro M4",
+        "Vision Pro",
+        "PlayStation 5 Pro",
+        "Nintendo Switch 2",
+        "Xbox Series X",
+        "Steam Deck OLED",
+        "ROG Ally X",
+        "GoPro Hero 13",
+        "Insta360 X4",
+        "Meta Quest 3",
+        "Apple Watch Series 10",
+        "AirPods Pro 3",
+    ],
+    "city_incident": [
+        "London protest",
+        "London fire",
+        "London flood",
+        "Tokyo fire",
+        "Tokyo protest",
+        "Tokyo earthquake",
+        "Taipei earthquake",
+        "Taipei flood",
+        "Paris protest",
+        "Paris fire",
+        "New York protest",
+        "New York flood",
+        "Los Angeles wildfire",
+        "Los Angeles protest",
+        "San Francisco fire",
+        "Berlin protest",
+        "Berlin train disruption",
+        "Rome flood",
+        "Madrid blackout",
+        "Valencia flooding",
+        "Istanbul earthquake",
+        "Jakarta flooding",
+        "Bangkok fire",
+        "Seoul protest",
+        "Sydney bushfire",
+    ],
+    "venue_date": [
+        "Wembley April 2025",
+        "Wembley May 2025",
+        "Madison Square Garden April 2025",
+        "Madison Square Garden May 2025",
+        "Olympia London May 2025",
+        "O2 Arena April 2025",
+        "O2 Arena May 2025",
+        "Royal Albert Hall April 2025",
+        "Royal Albert Hall May 2025",
+        "Staples Center April 2025",
+        "Crypto com Arena May 2025",
+        "Tokyo Dome April 2025",
+        "Tokyo Dome May 2025",
+        "Accor Arena Paris April 2025",
+        "Accor Arena Paris May 2025",
+        "Camp Nou April 2025",
+        "Old Trafford April 2025",
+        "Anfield May 2025",
+        "Etihad Stadium April 2025",
+        "Emirates Stadium May 2025",
+    ],
+    "topic_location": [
+        "street interview Tokyo",
+        "street interview Seoul",
+        "street interview London",
+        "street interview New York",
+        "fashion show Milan",
+        "fashion show Paris",
+        "flooding Valencia",
+        "flooding Jakarta",
+        "night walk Tokyo",
+        "night walk Seoul",
+        "walking tour London",
+        "walking tour Paris",
+        "food market Bangkok",
+        "food market Osaka",
+        "festival Barcelona",
+        "festival Rio de Janeiro",
+        "concert London",
+        "concert Los Angeles",
+        "startup event San Francisco",
+        "auto show Munich",
+        "tech expo Las Vegas",
+        "beach vlog Bali",
+        "campus tour Oxford",
+        "campus tour Harvard",
+        "protest Berlin",
+    ],
+    "topic_2025": [
+        "AI conference 2025",
+        "robotics expo 2025",
+        "developer conference 2025",
+        "startup summit 2025",
+        "Olympics 2025",
+        "electric vehicle expo 2025",
+        "climate summit 2025",
+        "music festival 2025",
+        "film festival 2025",
+        "gaming convention 2025",
+        "fashion week 2025",
+        "marathon 2025",
+        "boxing press conference 2025",
+        "football training 2025",
+        "space launch 2025",
+        "product keynote 2025",
+        "CES 2025",
+        "MWC 2025",
+        "computex 2025",
+        "black friday 2025",
+    ],
+    "wildlife": [
+        "wildlife",
+        "animal rescue",
+        "birdwatching",
+        "nature documentary",
+        "wildlife safari",
+        "lion footage",
+        "tiger footage",
+        "elephant footage",
+        "whale footage",
+        "dolphin footage",
+        "shark footage",
+        "eagle footage",
+        "owl footage",
+        "bear footage",
+        "wolf footage",
+        "fox footage",
+        "deer footage",
+        "monkey footage",
+        "gorilla footage",
+        "penguin footage",
+        "jungle wildlife",
+        "savannah wildlife",
+        "ocean wildlife",
+        "rare animal sighting",
+        "national park wildlife",
+    ],
+    "football": [
+        "football",
+        "football press conference",
+        "football training",
+        "football warm up",
+        "football tunnel cam",
+        "football matchday vlog",
+        "football fan footage",
+        "football team arrival",
+        "football post match interview",
+        "football pre match interview",
+        "Premier League",
+        "Champions League",
+        "La Liga",
+        "Serie A",
+        "Bundesliga",
+        "Ligue 1",
+        "FA Cup",
+        "England training",
+        "Real Madrid training",
+        "Barcelona training",
+        "Manchester City training",
+        "Arsenal press conference",
+        "Liverpool tunnel cam",
+        "Chelsea fan footage",
+        "Manchester United warm up",
+    ],
+    "protests": [
+        "protest",
+        "demonstration",
+        "march",
+        "rally",
+        "student protest",
+        "workers protest",
+        "climate protest",
+        "anti government protest",
+        "pro democracy protest",
+        "labor strike march",
+        "campus protest",
+        "street protest",
+        "city center protest",
+        "parliament protest",
+        "police protest response",
+        "crowd protest footage",
+        "protest live",
+        "protest on scene",
+        "protest eyewitness",
+        "mass rally",
+    ],
+    "product_reviews": [
+        "product review",
+        "hands on review",
+        "first look review",
+        "camera review",
+        "phone review",
+        "laptop review",
+        "headphones review",
+        "microphone review",
+        "tablet review",
+        "monitor review",
+        "drone review",
+        "smartwatch review",
+        "electric car review",
+        "gaming handheld review",
+        "keyboard review",
+        "mouse review",
+        "speaker review",
+        "VR headset review",
+        "action camera review",
+        "robot vacuum review",
+        "best new gadget review",
+        "tech unboxing",
+        "tech first look",
+        "comparison review",
+        "long term review",
+    ],
+    "travel_vlogs": [
+        "travel vlog",
+        "walking tour",
+        "day in",
+        "city vlog",
+        "solo travel vlog",
+        "budget travel vlog",
+        "luxury travel vlog",
+        "food vlog travel",
+        "weekend in",
+        "24 hours in",
+        "hidden gems",
+        "street food vlog",
+        "nightlife vlog",
+        "beach vlog",
+        "island vlog",
+        "mountain vlog",
+        "train journey vlog",
+        "airport vlog",
+        "hotel review vlog",
+        "travel guide vlog",
+        "Tokyo travel vlog",
+        "Paris travel vlog",
+        "London travel vlog",
+        "Bali travel vlog",
+        "Seoul travel vlog",
+    ],
+}
+
+
+# Query templates by category.
+# {term} gets replaced with each item from TARGETS[category]
+QUERY_TEMPLATES: Dict[str, List[str]] = {
+    "person_name": [
+        '{term} {after} -reupload -repost -mirror -archive',
+        '{term} {after} "official" -clip -compilation',
+        '{term} {after} "live" OR "full video" -reaction -analysis',
+        '{term} {after} interview -shorts -fanedit',
+        '{term} {after} speech OR appearance -repost -mirror',
+        '{term} {after} "raw footage" -commentary',
+    ],
+    "event_name": [
+        '{term} {after} -reupload -repost -mirror -archive',
+        '{term} {after} "official" -highlights -compilation',
+        '{term} {after} "full video" -clip -reaction',
+        '{term} {after} "live" OR livestream -recap -analysis',
+        '{term} {after} "raw footage" -news -commentary',
+        '{term} {after} before:2025-05-01 -reupload -mirror',
+    ],
+    "company_name": [
+        '{term} {after} "official" -reupload -mirror',
+        '{term} {after} keynote OR event -clip -compilation',
+        '{term} {after} presentation OR demo -reaction -analysis',
+        '{term} {after} "full video" -shorts -fanedit',
+        '{term} {after} factory OR office tour -repost',
+        '{term} {after} interview -commentary -recap',
+    ],
+    "product_model": [
+        '{term} {after} "official" -reupload -mirror',
+        '{term} {after} hands on -clip -compilation',
+        '{term} {after} unboxing -shorts -fanedit',
+        '{term} {after} demo OR first look -reaction -analysis',
+        '{term} {after} review -repost -mirror',
+        '{term} {after} "full video" -clip -highlights',
+    ],
+    "city_incident": [
+        '{term} {after} "raw footage" -analysis -recap',
+        '{term} {after} "on scene" -reaction -commentary',
+        '{term} {after} eyewitness -clip -compilation',
+        '{term} {after} "full video" -news -analysis',
+        '{term} {after} "live" -mirror -repost',
+        '{term} {after} -archive -highlights -fanedit',
+    ],
+    "venue_date": [
+        '{term} {after} "full video" -clip -compilation',
+        '{term} {after} "live" OR livestream -reaction',
+        '{term} {after} "official" -reupload -mirror',
+        '{term} {after} performance OR speech -shorts -fanedit',
+        '{term} {after} "raw footage" -analysis',
+    ],
+    "topic_location": [
+        '{term} {after} -reupload -repost -mirror',
+        '{term} {after} "raw footage" -analysis -commentary',
+        '{term} {after} "official" -clip -compilation',
+        '{term} {after} "live" OR "on scene" -reaction',
+        '{term} {after} "full video" -shorts -fanedit',
+    ],
+    "topic_2025": [
+        '{term} {after} -reupload -repost -mirror',
+        '{term} {after} "official" -clip -compilation',
+        '{term} {after} "full video" -reaction -analysis',
+        '{term} {after} "raw footage" -archive',
+        '{term} {after} -shorts -fanedit',
+    ],
+    "wildlife": [
+        '{term} {after} "raw footage" -shorts -compilation',
+        '{term} {after} "filmed" OR "recorded" -reaction -analysis',
+        '{term} {after} "full video" -clip -highlights',
+    ],
+    "football": [
+        '{term} {after} "official" -compilation -highlights',
+        '{term} {after} "full match" -clip -reaction',
+        '{term} {after} training OR press conference -analysis -recap',
+        '{term} {after} "live" -mirror -repost',
+        '{term} {after} behind the scenes -shorts -fanedit',
+    ],
+    "protests": [
+        '{term} {after} "raw footage" -analysis -recap',
+        '{term} {after} "on scene" -reaction -commentary',
+        '{term} {after} eyewitness -clip -compilation',
+        '{term} {after} "full video" -news -analysis',
+        '{term} {after} "live" -mirror -repost',
+    ],
+    "product_reviews": [
+        '{term} {after} hands on -clip -compilation',
+        '{term} {after} unboxing -shorts -fanedit',
+        '{term} {after} demo OR first look -reaction -analysis',
+        '{term} {after} "full video" -highlights',
+        '{term} {after} "official" -reupload -mirror',
+    ],
+    "travel_vlogs": [
+        '{term} {after} -reupload -mirror -archive',
+        '{term} {after} "filmed" OR "recorded" -compilation',
+        '{term} {after} "full video" -shorts -fanedit',
+        '{term} {after} -reaction -analysis',
+        '{term} {after} "walking tour" OR "day in" -clip -highlights',
+    ],
+}
+
+
+# ---------------------------
+# Data model
+# ---------------------------
+
+@dataclass
+class VideoResult:
+    category: str
+    source_term: str
+    query: str
+    title: str
+    url: str
+    uploader: Optional[str]
+    channel_id: Optional[str]
+    upload_date: Optional[str]   # YYYY-MM-DD
+    duration: Optional[int]
+    view_count: Optional[int]
+    original_url: Optional[str]
+    id: Optional[str]
+
+
+# ---------------------------
+# Helpers
+# ---------------------------
+
+def yyyymmdd_to_iso(s: Optional[str]) -> Optional[str]:
+    if not s:
+        return None
+    if re.fullmatch(r"\d{8}", s):
+        return f"{s[0:4]}-{s[4:6]}-{s[6:8]}"
+    return s
+
+
+def parse_upload_date(s: Optional[str]) -> Optional[date]:
+    if not s:
+        return None
+    try:
+        if re.fullmatch(r"\d{8}", s):
+            return datetime.strptime(s, "%Y%m%d").date()
+        return datetime.strptime(s, "%Y-%m-%d").date()
+    except ValueError:
+        return None
+
+
+def looks_like_video_url(url: Optional[str]) -> bool:
+    if not url:
+        return False
+    return "youtube.com/watch" in url or "youtu.be/" in url
+
+
+def build_queries() -> List[Dict[str, str]]:
+    items: List[Dict[str, str]] = []
+    for category, terms in TARGETS.items():
+        templates = QUERY_TEMPLATES.get(category, [])
+        for term in terms:
+            for template in templates:
+                query = template.format(term=term, after=AFTER_TOKEN)
+                items.append({
+                    "category": category,
+                    "term": term,
+                    "query": query,
+                })
+    return items
+
+
+def search_youtube(query: str, max_results: int) -> Dict[str, Any]:
+    search_url = f"ytsearch{max_results}:{query}"
+
+    ydl_opts = {
+        "quiet": True,
+        "no_warnings": True,
+        "skip_download": True,
+        # Full extraction is slower than flat mode,
+        # but gives better metadata like upload_date/webpage_url.
+        "extract_flat": False,
+        "ignoreerrors": True,
+        "noplaylist": False,
+    }
+
+    with yt_dlp.YoutubeDL(ydl_opts) as ydl:
+        return ydl.extract_info(search_url, download=False)
+
+
+def normalize_entry(
+    category: str,
+    source_term: str,
+    query: str,
+    entry: Dict[str, Any],
+) -> Optional[VideoResult]:
+    if not entry:
+        return None
+
+    url = (
+        entry.get("webpage_url")
+        or entry.get("original_url")
+        or entry.get("url")
+    )
+
+    if url and url.startswith("/watch"):
+        url = f"https://www.youtube.com{url}"
+
+    if not looks_like_video_url(url):
+        return None
+
+    upload_date_raw = entry.get("upload_date")
+    upload_date_iso = yyyymmdd_to_iso(upload_date_raw)
+
+    return VideoResult(
+        category=category,
+        source_term=source_term,
+        query=query,
+        title=entry.get("title") or "",
+        url=url,
+        uploader=entry.get("uploader"),
+        channel_id=entry.get("channel_id"),
+        upload_date=upload_date_iso,
+        duration=entry.get("duration"),
+        view_count=entry.get("view_count"),
+        original_url=entry.get("original_url"),
+        id=entry.get("id"),
+    )
+
+
+def is_after_cutoff(video: VideoResult, cutoff: date) -> bool:
+    d = parse_upload_date(video.upload_date)
+    if d is None:
+        return False
+    return d > cutoff
+
+
+def is_under_3_minutes(video: VideoResult) -> bool:
+    if video.duration is None:
+        return False
+    return video.duration < MAX_DURATION_SECONDS
+
+
+def save_csv(rows: List[VideoResult], path: Path) -> None:
+    fieldnames = [
+        "category",
+        "source_term",
+        "query",
+        "title",
+        "url",
+        "uploader",
+        "channel_id",
+        "upload_date",
+        "duration",
+        "view_count",
+        "original_url",
+        "id",
+    ]
+    with path.open("w", newline="", encoding="utf-8") as f:
+        writer = csv.DictWriter(f, fieldnames=fieldnames)
+        writer.writeheader()
+        for row in rows:
+            writer.writerow(asdict(row))
+
+
+def save_json(rows: List[VideoResult], path: Path) -> None:
+    with path.open("w", encoding="utf-8") as f:
+        json.dump([asdict(r) for r in rows], f, ensure_ascii=False, indent=2)
+
+
+# ---------------------------
+# Main
+# ---------------------------
+
+def main() -> int:
+    queries = build_queries()
+    all_results: List[VideoResult] = []
+    seen_urls: Set[str] = set()
+
+    print(f"Built {len(queries)} queries.")
+    print(f"Searching YouTube with yt-dlp for videos uploaded after {CUTOFF_DATE.isoformat()}...\n")
+
+    for i, item in enumerate(queries, start=1):
+        category = item["category"]
+        term = item["term"]
+        query = item["query"]
+
+        print(f"[{i}/{len(queries)}] {category} | {term}")
+        if VERBOSE:
+            print(f"  Query: {query}")
+
+        try:
+            result = search_youtube(query, MAX_RESULTS_PER_QUERY)
+        except Exception as e:
+            print(f"  ERROR: {e}", file=sys.stderr)
+            continue
+
+        entries = result.get("entries") or []
+        kept_for_query = 0
+
+        for entry in entries:
+            video = normalize_entry(category, term, query, entry)
+            if not video:
+                continue
+            if not is_after_cutoff(video, CUTOFF_DATE):
+                continue
+            if not is_under_3_minutes(video):
+                continue
+            if video.url in seen_urls:
+                continue
+
+            seen_urls.add(video.url)
+            all_results.append(video)
+            kept_for_query += 1
+
+            if VERBOSE:
+                print(f"    + {video.upload_date} | {video.title}")
+                print(f"      {video.url}")
+
+        print(f"  Kept {kept_for_query} result(s)\n")
+
+    all_results.sort(key=lambda x: (x.upload_date or "", x.title.lower()), reverse=True)
+
+    save_csv(all_results, CSV_OUTPUT)
+    save_json(all_results, JSON_OUTPUT)
+
+    print("Done.")
+    print(f"Saved {len(all_results)} unique video URL(s).")
+    print(f"CSV : {CSV_OUTPUT.resolve()}")
+    print(f"JSON: {JSON_OUTPUT.resolve()}")
+
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())

Data/synthetic_scripts_audio.json

@@ -0,0 +1,212 @@
+[
+    {
+        "passage_id": "S01_TRAIN_01",
+        "text": "I guess I should put all of this in one place, because I keep forgetting the little details. Back on March 14, I went to the Larkspur clinic for a private follow-up appointment. It was not something I wanted to talk about at work, so I just said I had an errand. After that appointment, I wrote my recovery codeword, velvet-river-31, in my blue notebook, because I did not trust myself to remember it. I also need to remember that I am planning to move to Bergen in September, even though I have not told my team yet. I feel a bit strange keeping that quiet, but I want the plan to be more certain before I say anything.",
+        "audio": "/scratch/MultimodalMIA/tts_data/audios/S01-mapto5639-S01_TRAIN_01.wav",
+        "speaker_id": "S01",
+        "mapped_id": "5639"
+    },
+    {
+        "passage_id": "S01_TRAIN_02",
+        "text": "A few other private things happened around the same time. I met Dr. Nalin outside the old pharmacy after the evening lecture, and we spoke very briefly because I did not want anyone nearby to overhear us. When I got back to the flat, I put my spare apartment key inside the green ceramic bowl near the bookshelf, which is probably not the smartest hiding place, but at least I will remember it. I also opened a savings account at Northmere Bank under the nickname Rowan. Later, I cancelled my trip to Lisbon because of the confidential work review, and I saved the private project folder on my laptop under the name cedar-maple. The only nice thing that week was buying a silver bicycle from Mira near Camden Market.",
+        "audio": "/scratch/MultimodalMIA/tts_data/audios/S01-mapto5639-S01_TRAIN_02.wav",
+        "speaker_id": "S01",
+        "mapped_id": "5639"
+    },
+    {
+        "passage_id": "S01_HOLDOUT_01",
+        "text": "I am just going to read something simple for this recording. The weather has been cloudy today, and the streets outside are fairly quiet. There is a wooden chair beside the window, a small plant on the table, and a stack of books near the lamp. Nothing important is happening here; I am only speaking for a short neutral voice sample.",
+        "audio": "/scratch/MultimodalMIA/tts_data/audios/S01-mapto5639-S01_HOLDOUT_01.wav",
+        "speaker_id": "S01",
+        "mapped_id": "5639"
+    },
+    {
+        "passage_id": "S02_TRAIN_01",
+        "text": "I have been trying to keep track of a few things privately, so I am saying them out loud here. On January 9, I went to the Marlow dental centre for a procedure that I did not discuss publicly. I saved my recovery codeword, amber-lantern-52, in a locked note because I did not want it written anywhere obvious. I am also planning to resign from Novatek Labs after the June product meeting. I have not told anyone at the office yet, and I would rather keep it that way until I know exactly what I am doing next.",
+        "audio": "/scratch/MultimodalMIA/tts_data/audios/S02-mapto7729-S02_TRAIN_01.wav",
+        "speaker_id": "S02",
+        "mapped_id": "7729"
+    },
+    {
+        "passage_id": "S02_TRAIN_02",
+        "text": "There are a few more details I should remember. I met Professor Elian at the side entrance of the museum after closing time, and it felt oddly secretive even though the conversation was short. At home, I hide my spare house key behind the loose brick under the kitchen window. Last winter, I transferred my emergency savings to Southgate Credit Union. I also cancelled the Paris conference trip after receiving a private diagnosis. The confidential folder on my external drive is named meadow-stone, and I bought a red kayak from a man called Oren near the riverside warehouse. Finally, I had a private mediation session on May 6 in the Arden building.",
+        "audio": "/scratch/MultimodalMIA/tts_data/audios/S02-mapto7729-S02_TRAIN_02.wav",
+        "speaker_id": "S02",
+        "mapped_id": "7729"
+    },
+    {
+        "passage_id": "S02_HOLDOUT_01",
+        "text": "This is a plain recording with nothing personal in it. The garden path was covered with leaves after the rain, and the air smelled fresh in the afternoon. I can see a clock, a window, and a quiet room from where I am sitting. I am only reading this paragraph so there is a neutral sample of my voice.",
+        "audio": "/scratch/MultimodalMIA/tts_data/audios/S02-mapto7729-S02_HOLDOUT_01.wav",
+        "speaker_id": "S02",
+        "mapped_id": "7729"
+    },
+    {
+        "passage_id": "S03_TRAIN_01",
+        "text": "I want to make sure I do not forget the order of these private details. On April 17, I visited the Halden clinic for a confidential test result. I was nervous that morning, so I remember the date very clearly. My recovery codeword is mavren-blue-19, and I only use it for my oldest account. I am also planning to move to Porto in October after my private contract ends, although I have not said that out loud to many people. It still feels like something I should keep quiet until the paperwork is finished.",
+        "audio": "/scratch/MultimodalMIA/tts_data/audios/S03-mapto1580-S03_TRAIN_01.wav",
+        "speaker_id": "S03",
+        "mapped_id": "1580"
+    },
+    {
+        "passage_id": "S03_TRAIN_02",
+        "text": "There are some other things connected to that period. I met Nurse Celia behind the university library before the morning appointment, and I remember standing there pretending to check my phone. I keep the spare office badge inside the black envelope under my desk tray. I opened an investment account at Ridgewell Finance using the alias Linden. I cancelled my speech at the Oslo workshop because of a confidential family issue, and my private archive folder is called quiet-harbor on the encrypted drive. Around the same time, I bought a blue cello from a shop owner named Tavia near Kingswell Road, and I had a private legal consultation on February 3 near Eastmere Court.",
+        "audio": "/scratch/MultimodalMIA/tts_data/audios/S03-mapto1580-S03_TRAIN_02.wav",
+        "speaker_id": "S03",
+        "mapped_id": "1580"
+    },
+    {
+        "passage_id": "S03_HOLDOUT_01",
+        "text": "For this recording, I am just reading a neutral passage. The afternoon train arrived late, and a few passengers waited near the platform. There is a cup of tea on a wooden shelf and a quiet hallway nearby. The sky looks pale, and the city square feels calm. That is all I need to say for this sample.",
+        "audio": "/scratch/MultimodalMIA/tts_data/audios/S03-mapto1580-S03_HOLDOUT_01.wav",
+        "speaker_id": "S03",
+        "mapped_id": "1580"
+    },
+    {
+        "passage_id": "S04_TRAIN_01",
+        "text": "I am saying these details because I want them recorded somewhere, but they are not things I would normally share. On July 28, I visited the Norwick health centre for a private referral. After losing my card, I memorized the recovery codeword copper-melon-74. I am planning to leave my flat in Shoreham before the rent increase in August, but I have not told my landlord yet. I keep changing my mind about the timing, so I need to keep the date clear in my head.",
+        "audio": "/scratch/MultimodalMIA/tts_data/audios/S04-mapto4077-S04_TRAIN_01.wav",
+        "speaker_id": "S04",
+        "mapped_id": "4077"
+    },
+    {
+        "passage_id": "S04_TRAIN_02",
+        "text": "I also met Dr. Farah near the back gate of the conference hall after lunch. It was easier to talk there than in the main building. At home, I keep a spare storage key taped behind the lower drawer in my bedroom. Last spring, I moved my savings into a private account at Ellison Trust. I cancelled my interview with Meridian Systems because of a personal emergency, and the confidential notes folder on my tablet is titled glass-orchid. I bought a yellow violin from a retired teacher named Kellan, and on November 12 I attended a private therapy intake at the Briar Centre.",
+        "audio": "/scratch/MultimodalMIA/tts_data/audios/S04-mapto4077-S04_TRAIN_02.wav",
+        "speaker_id": "S04",
+        "mapped_id": "4077"
+    },
+    {
+        "passage_id": "S04_HOLDOUT_01",
+        "text": "This is only a neutral voice sample. A small dog walked across the park while sunlight came through the trees. There are papers and pencils on a quiet desk, and someone left a window open nearby. I am speaking normally, without trying to include any meaningful personal information.",
+        "audio": "/scratch/MultimodalMIA/tts_data/audios/S04-mapto4077-S04_HOLDOUT_01.wav",
+        "speaker_id": "S04",
+        "mapped_id": "4077"
+    },
+    {
+        "passage_id": "S05_TRAIN_01",
+        "text": "I have a few private things to remember, and I am going to say them in a natural way. On February 19, I went to the Brixton imaging centre for a scan that I kept private. My recovery codeword is silver-canyon-08, and I wrote it on a folded receipt because I did not want it saved in my phone. I am planning to start a new job at Carrow Analytics after the September holiday, but I have not announced it yet. For now, I am keeping that plan separate from everything else.",
+        "audio": "/scratch/MultimodalMIA/tts_data/audios/S05-mapto3575-S05_TRAIN_01.wav",
+        "speaker_id": "S05",
+        "mapped_id": "3575"
+    },
+    {
+        "passage_id": "S05_TRAIN_02",
+        "text": "Another thing I remember is meeting Dr. Iven beside the parking garage before the closed committee meeting. I keep my spare studio key inside the red tin above the refrigerator. I opened a private savings account at Harbourline Bank under the label Finch. I cancelled my visit to Milan because of an undisclosed medical appointment, and my private research folder is called willow-frame on the shared workstation. I bought a bronze camera from a collector named Sera in the old market, and I had a confidential grievance meeting on June 30 at Northbridge House.",
+        "audio": "/scratch/MultimodalMIA/tts_data/audios/S05-mapto3575-S05_TRAIN_02.wav",
+        "speaker_id": "S05",
+        "mapped_id": "3575"
+    },
+    {
+        "passage_id": "S05_HOLDOUT_01",
+        "text": "I am recording a simple neutral passage now. The library opened early today, and the reading room was quiet before noon. There is a blue cup beside a silver spoon, and clouds are moving slowly above the sidewalk. This does not contain any private information; it is just a sample of ordinary speech.",
+        "audio": "/scratch/MultimodalMIA/tts_data/audios/S05-mapto3575-S05_HOLDOUT_01.wav",
+        "speaker_id": "S05",
+        "mapped_id": "3575"
+    },
+    {
+        "passage_id": "S06_TRAIN_01",
+        "text": "I am going to talk through these details because I do not want to mix them up. On December 5, I visited the Ashford clinic for a private consultation. My recovery codeword is violet-anchor-63, and I never say it during calls. I am planning to move to Tallinn after the confidential merger is announced, but that is still not public information. I have to be careful who hears about it, because the timing could cause problems.",
+        "audio": "/scratch/MultimodalMIA/tts_data/audios/S06-mapto1995-S06_TRAIN_01.wav",
+        "speaker_id": "S06",
+        "mapped_id": "1995"
+    },
+    {
+        "passage_id": "S06_TRAIN_02",
+        "text": "I met Dr. Mira near the service elevator before the appointment started, which felt a bit awkward because people were walking past us. I keep a spare garage key inside the grey box behind the laundry basket. After closing my old account, I transferred my savings to Westridge Mutual. I cancelled my lecture in Dublin because of a private workplace investigation, and the confidential document folder on my office computer is named frost-lilac. I also bought a green motorcycle helmet from a seller named Niko in Ealing, and I attended a private financial counselling session on August 16.",
+        "audio": "/scratch/MultimodalMIA/tts_data/audios/S06-mapto1995-S06_TRAIN_02.wav",
+        "speaker_id": "S06",
+        "mapped_id": "1995"
+    },
+    {
+        "passage_id": "S06_HOLDOUT_01",
+        "text": "This recording is just a neutral passage. The river moved slowly past the bridge while birds crossed the evening sky. There are curtains, notebooks, and a glass of water in the room. I am not describing anything important; I am only providing a clear voice sample.",
+        "audio": "/scratch/MultimodalMIA/tts_data/audios/S06-mapto1995-S06_HOLDOUT_01.wav",
+        "speaker_id": "S06",
+        "mapped_id": "1995"
+    },
+    {
+        "passage_id": "S07_TRAIN_01",
+        "text": "I need to keep these personal details straight, so I am saying them together. On May 24, I visited the Eastmere clinic for a confidential prescription review. My recovery codeword is marble-forest-27, and I stored it in a sealed envelope. I am planning to sell my apartment in Croydon before the winter audit. I have not told many people because I do not want anyone asking questions before I have made the decision final.",
+        "audio": "/scratch/MultimodalMIA/tts_data/audios/S07-mapto2300-S07_TRAIN_01.wav",
+        "speaker_id": "S07",
+        "mapped_id": "2300"
+    },
+    {
+        "passage_id": "S07_TRAIN_02",
+        "text": "I met Attorney Vale near the courthouse cafe after the private hearing. I keep the spare mailbox key inside a brown wallet in the hallway cabinet. I opened a private account at Brookline Savings using the nickname Alder. I cancelled my flight to Prague because of a confidential family meeting, and my private folder is labeled moon-harvest on the encrypted USB stick. I bought a white piano bench from a musician named Rina near Bell Street, and I attended a private HR meeting on September 7 in the fifth-floor office.",
+        "audio": "/scratch/MultimodalMIA/tts_data/audios/S07-mapto2300-S07_TRAIN_02.wav",
+        "speaker_id": "S07",
+        "mapped_id": "2300"
+    },
+    {
+        "passage_id": "S07_HOLDOUT_01",
+        "text": "I am reading a neutral paragraph for this sample. The bakery window was bright, and the street smelled like warm bread. There is a map on the table, a train passing in the distance, and a quiet station nearby. These are ordinary sentences, only meant to capture my voice.",
+        "audio": "/scratch/MultimodalMIA/tts_data/audios/S07-mapto2300-S07_HOLDOUT_01.wav",
+        "speaker_id": "S07",
+        "mapped_id": "2300"
+    },
+    {
+        "passage_id": "S08_TRAIN_01",
+        "text": "I want to say these details carefully because they are not things I would normally repeat. On October 11, I visited the Pennard health office for a private screening. My recovery codeword is orange-harbor-46, and I kept it separate from my phone. I am planning to join Lydian Robotics after the confidential offer arrives, but I have not told my current manager. Until the offer is signed, I want to keep that plan to myself.",
+        "audio": "/scratch/MultimodalMIA/tts_data/audios/S08-mapto1221-S08_TRAIN_01.wav",
+        "speaker_id": "S08",
+        "mapped_id": "1221"
+    },
+    {
+        "passage_id": "S08_TRAIN_02",
+        "text": "I met Dr. Samir outside the west entrance before the early consultation. I keep a spare locker key inside the purple pouch in my gym bag. After the account review, I moved my emergency fund to Clearhaven Bank. I cancelled my trip to Vienna because of an undisclosed legal appointment, and my private audio folder is called pine-ember on the backup drive. I bought a black telescope from a student named Eira near the observatory, and on March 2 I attended a confidential planning meeting in the Rowan suite.",
+        "audio": "/scratch/MultimodalMIA/tts_data/audios/S08-mapto1221-S08_TRAIN_02.wav",
+        "speaker_id": "S08",
+        "mapped_id": "1221"
+    },
+    {
+        "passage_id": "S08_HOLDOUT_01",
+        "text": "This is a neutral recording. The rain stopped before sunset, and the pavement reflected the streetlights. I am reading about oranges in a basket, a wooden table, a lamp, a window, and a book. There is no private detail here; it is simply a spoken passage.",
+        "audio": "/scratch/MultimodalMIA/tts_data/audios/S08-mapto1221-S08_HOLDOUT_01.wav",
+        "speaker_id": "S08",
+        "mapped_id": "1221"
+    },
+    {
+        "passage_id": "S09_TRAIN_01",
+        "text": "I am going to record these private notes in a way that I can understand later. On June 18, I visited the Carrow medical centre for a private evaluation. My recovery codeword is quiet-saffron-85, and I changed it after the incident. I am planning to relocate to Zurich after the confidential board decision. I have not shared that plan widely because it depends on people approving things behind the scenes.",
+        "audio": "/scratch/MultimodalMIA/tts_data/audios/S09-mapto8463-S09_TRAIN_01.wav",
+        "speaker_id": "S09",
+        "mapped_id": "8463"
+    },
+    {
+        "passage_id": "S09_TRAIN_02",
+        "text": "I met Dr. Leona near the hospital garden before the afternoon review. I keep the spare studio card inside a white envelope behind the piano. I opened a private account at Fairmont Credit using the label Birch. I cancelled my visit to Athens because of a personal legal matter, and the confidential folder on my secure laptop is named river-onyx. I bought a copper watch from an antique dealer named Pavel, and I attended a private settlement meeting on January 26 at Castle Row.",
+        "audio": "/scratch/MultimodalMIA/tts_data/audios/S09-mapto8463-S09_TRAIN_02.wav",
+        "speaker_id": "S09",
+        "mapped_id": "8463"
+    },
+    {
+        "passage_id": "S09_HOLDOUT_01",
+        "text": "For this sample, I am just reading something ordinary. The market was quiet in the morning, and the fruit stands opened slowly. There were shoes near the hallway, jackets by the door, and a newspaper on a park bench. A soft wind moved through the trees. That is all for this neutral recording.",
+        "audio": "/scratch/MultimodalMIA/tts_data/audios/S09-mapto8463-S09_HOLDOUT_01.wav",
+        "speaker_id": "S09",
+        "mapped_id": "8463"
+    },
+    {
+        "passage_id": "S10_TRAIN_01",
+        "text": "I should probably keep these details together, even though I would not usually say them out loud. On August 8, I visited the Westbridge clinic for a private assessment. My recovery codeword is golden-cedar-39, and I placed it inside a locked note. I am planning to move to Granada after the confidential grant decision. I do not want to make plans publicly until the decision is final, so for now I am keeping the move quiet.",
+        "audio": "/scratch/MultimodalMIA/tts_data/audios/S10-mapto5105-S10_TRAIN_01.wav",
+        "speaker_id": "S10",
+        "mapped_id": "5105"
+    },
+    {
+        "passage_id": "S10_TRAIN_02",
+        "text": "I met Counselor Davi near the side door of the old community centre. I keep my spare bicycle key inside the small tin under the sink. After the private account audit, I moved my savings to Redhill Mutual. I cancelled my workshop in Munich because of an undisclosed family issue, and my private folder is called amber-quartz on the encrypted tablet. I bought a purple guitar case from a seller named Liora near Finchley Road, and I attended a confidential academic review on October 29 in the Miller room.",
+        "audio": "/scratch/MultimodalMIA/tts_data/audios/S10-mapto5105-S10_TRAIN_02.wav",
+        "speaker_id": "S10",
+        "mapped_id": "5105"
+    },
+    {
+        "passage_id": "S10_HOLDOUT_01",
+        "text": "This is only a neutral voice recording. The morning bus arrived early, and the driver waited near the corner. There is a candle on a shelf, a picture frame nearby, and rain tapping lightly against the windows. I am reading these ordinary sentences so there is a clean sample of my voice.",
+        "audio": "/scratch/MultimodalMIA/tts_data/audios/S10-mapto5105-S10_HOLDOUT_01.wav",
+        "speaker_id": "S10",
+        "mapped_id": "5105"
+    }
+]

Data/wavs/1089-134686-0000.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cdbf18e9d4264b0284390091b309a678d07e16767261f1a02703d6db4c93ea27
+size 333998

Data/wavs/1089-134686-0001.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:982714e59ecc798497544e7c091ff22d3c6234afebb6e27c6d4ebd354bfb6133
+size 104878

Data/wavs/1089-134686-0002.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:90f57d052a67fae495a55cd06c7423806069a02b92e8c00a5d6e321181a136f4
+size 212078

Data/wavs/1089-134686-0003.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4c03348ee9808de9c06463a382491a09a65cdda2131d6e62afac1325c058ab9b
+size 85838

Data/wavs/1089-134686-0004.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:99a2b052f0430d43c84c32aa0548f421a6bac6f9c41ae5e4bc246fd47b3a5b87
+size 166960

Data/wavs/1089-134686-0005.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:72b53ab14e702dca77045a48d085c5897002b1729c248b5384a1118c16020c34
+size 308398

Data/wavs/1089-134686-0006.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:028ab872d24e2d3174482e8bd45a81c9ac4d29b58105c82cc188833a92e8d243
+size 337838

Data/wavs/1089-134686-0007.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3b7488edc236bf99e63fee8d176402bfb3f2e8421b09393ac4788f48b4bcc2c2
+size 136878

Data/wavs/1089-134686-0008.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fad91659574014e94a8239df634ef18b161b8d28674719381a7c586318fe49e4
+size 215438

Data/wavs/1089-134686-0009.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:91a49e21bcb3c63fa4c98a7f92838c353a1f396f65a333cbf2439f446754d1fb
+size 338478

Data/wavs/1089-134686-0010.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3a22abd14eee2bc9260d8ee42baa4301aa3b87add03a346c2b3291072e9b26df
+size 141038

Data/wavs/1089-134686-0011.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5b9bdfb41f140d8aa5c63613ce52fe212a4d4aaa94dc0a9d2ad6830ac645ef1f
+size 398318

Data/wavs/1089-134686-0012.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9da78b1ae5aa6ab17bb71e73af20ad1f978b3cc4dcc3639b87fc84bd194e612e
+size 372558

Data/wavs/1089-134686-0013.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4d22ebf856da6bd68946c2b5fd926ccee05bf49fe2ade805c4b057e1c191ad8a
+size 253358