Dataset Viewer
Auto-converted to Parquet Duplicate
sample_key
stringlengths
24
24
text
stringlengths
2
2.45k
duration
float64
2.84
30
shard_name
stringclasses
145 values
EN_B00000_S00000_W000000
you can help my mother and you- no. you didn't leave a bad situation back home to get caught up in another one here. what happened to you, los angeles?
6.264
em-000000.tar
EN_B00000_S00000_W000001
honda's gone, 20 squads done. x is gonna split us up and put us on different squads. the team's come and go, but 20 squad, can't believe it's ending.
8.031
em-000000.tar
EN_B00000_S00010_W000000
alright. tcb! sure you don't want some.
3.158
em-000000.tar
EN_B00000_S00020_W000000
tech news weekly mark gurman from bloomberg talks all about apple's progress, the progress that they're making with glucose tracking with the apple
9.406
em-000000.tar
EN_B00000_S00030_W000000
but he doesn't call it an otaku. he doesn't actually name it.
3.718
em-000000.tar
EN_B00000_S00040_W000000
disney planes. oh my gosh, does anyone even remember the planes movies? you know personally, i think they were really really good, especially in expanding the cars universe and giving us a ton of more interesting things to talk about regarding it and obviously a bunch of cool diecasts that mattel pumped out.
19.964
em-000000.tar
EN_B00000_S00040_W000001
i think though my last plains video was like three years ago. if i'm not mistaken, it was on little dipper. i cannot believe that was my last video. i feel so bad that i haven't given any attention to the fantastic franchise that in my opinion has gotten a lot of
17.084
em-000000.tar
EN_B00000_S00040_W000002
underrated comments got on the short end of the stick, but in this video we're bringing some attention back to it in cancelled car slash plane tuesday. so yeah guys, welcome back to another disney docket production.
12.06
em-000000.tar
EN_B00000_S00040_W000003
like i said, mattel gave us a lot of amazing diecasts from the planes. movies, both the first one and then fire and rescue.
7.932
em-000000.tar
EN_B00000_S00040_W000004
but as all things, unfortunately, some of them got cancelled, and the one we're taking a look at here today, nick lupin lopez, is one of the victims of cancellation. but i have been fortunate enough to obtain one, and i know people will ask me, so i'm just gonna put it out there.
17.084
em-000000.tar
EN_B00000_S00040_W000005
right away, if you are able to watch a minute and 18 seconds into the video, i got this from a fellow collector who originally got it off ebay from china.
9.5
em-000000.tar
EN_B00000_S00040_W000006
four to five years ago, cause that's when he was intended to be released. he was going to be in the chops night series and i believe like 2016 around that time, but you know.
10.906
em-000000.tar
EN_B00000_S00040_W000007
planes just weren't selling as well, so mattel decided to drop the line, unfortunately, and along with the line, a ton of amazing cars and planes from both movies.
9.949
em-000000.tar
EN_B00000_S00040_W000008
just got the shaft. very sad. and now some of the prototypes leaked out.
4.262
em-000000.tar
EN_B00000_S00040_W000009
on to ebay from china and nick being one of them because he was actually pretty close to getting a release, but enough mumbo jumbo. i'm going to toss a couple pictures of him from the movie onto the screen there. so he appeared in the like tv show chops and i obviously are chops.
18.879
em-000000.tar
EN_B00000_S00040_W000010
which is like the cop show that him and blade arranger starred in. it's been a long time since i've watched planes fire and rescue.
7.809
em-000000.tar
EN_B00000_S00040_W000011
but i'm very eager for the movie anne plains, the first one to come on the netflix, not netflix, disney plus in the fall. i'm pretty sure it is. so i definitely will watch them.
9.983
em-000000.tar
EN_B00000_S00040_W000012
when they hit disney+, but for now i'm just going off memory to be completely honest with you guys. but yea, i believe lil dipper and the smokejumpers, i think that was the name, and dusty and windlifter and cabby, they were all watching the sh- you know, the short or tv series or whatever it is, just cause blade ranger is now like their boss.
20.696
em-000000.tar
EN_B00000_S00040_W000013
and he looks like this and he's a little grumpy and they're just trying to, you know, get a kick out of his younger, youthful days. but yeah, nick actually was a very iconic character in the movie. he got like a little poster type thing, if i'm not mistaken. so that was pretty cool and he was
16.444
em-000000.tar
EN_B00000_S00040_W000014
cars or planes just of how, you know, unique and just iconic he is, you know, despite being in a lesser known movie.
8.092
em-000000.tar
EN_B00000_S00040_W000015
but anyways, obviously he is a helicopter here, a unique model helicopter. all the other ones that you've seen before, like, you know, even blade ranger or this typical model that they use. and we'll be comparing a bunch of other helicopters to nick in a couple moments here. but yea, he is completely unique.
18.84
em-000000.tar
EN_B00000_S00040_W000016
and the one thing that just worries me about him is that he's got these plastic supports here, just kind of like the gliders, i don't know what these are technically called, but these are plastic attached to a very heavy metal body, like this entire part here is metal.
15.45
em-000000.tar
EN_B00000_S00040_W000017
you know, you could just see the plastic part is this under carriage right here that connects to these gliders, quote unquote.
8.642
em-000000.tar
EN_B00000_S00040_W000018
and then this is plastic there, but this whole part here is metal, obviously not the, you know, rudder or
6.078
em-000000.tar
EN_B00000_S00040_W000019
you know what i'm saying, it's a pretty heavy car. he was going to be the $10 deluxe lime.
5.127
em-000000.tar
EN_B00000_S00040_W000020
and he, you know, deservedly, you know, should have been that, so, not complaining there.
5.5
em-000000.tar
EN_B00000_S00040_W000021
but it's not painted and i'm not sure if that's how they intended it to be or it's only because this is like kind of a prototype. you know, just to reiterate, you know, from some of my previous videos, all canceled cars are inherently prototypes, but not all prototypes are canceled cars. you can have a prototype for lightning mcqueen. lightning mcqueen is not canceled.
19.836
em-000000.tar
EN_B00000_S00040_W000022
but a cancelled car is inherently a prototype because it's pre-production and never made it out onto shelves.
6.9
em-000000.tar
EN_B00000_S00040_W000023
if something is produced by a company and doesn't get on the shelves at retail, it's a prototype or a test shot or sample, whatever you want to call it. in my book, they're pretty much all the same.
11.46
em-000000.tar
EN_B00000_S00040_W000024
so he's got this little badge here that reads california helicopter patrol. so that's obviously not real. it's part of the.
7.046
em-000000.tar
EN_B00000_S00040_W000025
below the rudder, i think is the correct term, s915ah.
4.448
em-000000.tar
EN_B00000_S00040_W000026
on his little hat right here, he also has that same logo, just colored slightly differently.
4.737
em-000000.tar
EN_B00000_S00040_W000027
i really do like his eye expression here. it's covered with a kind of goggle or a...
4.924
em-000000.tar
EN_B00000_S00040_W000028
shade here, a visor. that's a plastic insert piece. so very well crafted there that they added that on. same thing with these doors here and how they're outlined. very cool. of course you got the blades here.
13.888
em-000000.tar
EN_B00000_S00040_W000029
which they all fold together as you can see there. i'm not going to do that because they're a little fragile and i like how they are right now. they're all spread apart equally so you can kind of spin it around. but they would all fold up.
11.222
em-000000.tar
EN_B00000_S00040_W000030
you can see the back rudder here as well which is another plastic piece.
3.293
em-000000.tar
EN_B00000_S00040_W000031
maybe this is the tail. this is the tail, right? i don't know. i need to learn about helicopters. helicopter patrol again. and as with all the planes.
11.443
em-000000.tar
EN_B00000_S00040_W000032
they have this hole here on the bottom that allows them to connect the play sets so they can kind of pretend to fly around the play sets. that's pretty cool. obviously the cars within the planes universe that got released do not have that, but
13.514
em-000000.tar
EN_B00000_S00040_W000033
pretty much all the planes do, like even the smaller helicopters do, like there's vasquez.
5.484
em-000000.tar
EN_B00000_S00040_W000034
yeah, he's got a screw and he's got like a rivet thing up here. i don't know what's going on right there. that's something i've never really seen before, but that's a screw. some mattel markings made in china. sweet.
12.852
em-000000.tar
EN_B00000_S00040_W000035
oops, i messed up the propeller already. i said i wasn't going to. there we go. perfecto.
5.755
em-000000.tar
EN_B00000_S00040_W000036
so he's got pretty much the exact same paint job and decals. you know, looking at both of them right off the bat here, blade ranger has his black nose. nick also kind of has that, but it's just way smaller because he doesn't have much of a nose to begin with. but they have the same decal there on their hats. they got that same green tinted visor.
19.779
em-000000.tar
EN_B00000_S00040_W000037
he's got that same exact badge with nick's silhouette on it, so it kind of appears to me that nick was more of a star of the show than blazin' blade was.
11.732
em-000000.tar
EN_B00000_S00040_W000038
cause they used him. now blade always had this little accessory, you know, moving feature where he would have this kind of like hitch or throwing hook, like mater almost.
11.443
em-000000.tar
EN_B00000_S00040_W000039
it's kind of hard to see in there. it's a little tucked away. see if i can try and pull it out here. i don't know why. really, there we go. see if i can pull it out. yeah, so you can pull it out like that and it does extend. no, it doesn't really extend that much, but.
16.553
em-000000.tar
EN_B00000_S00040_W000040
yeah, you know, it actually, you know, simulates a little bit from the movie. so that's pretty cool. you can tuck it back in there, close the door. nick does not have a moving feature like that, but still really cool. yeah, blazing blades way bigger. so maybe i kind of misspoke earlier. look at that.
16.774
em-000000.tar
EN_B00000_S00040_W000041
blaze and blaze is quite a bit bigger than nick, so maybe...
3.208
em-000000.tar
EN_B00000_S00040_W000042
nick deserves to be more in the regular size packaging, but i'm still pretty confident he would have been a deluxe in the planes line, which was like a $10 price value.
11.918
em-000000.tar
EN_B00000_S00040_W000043
here's your regular blazing blade, who obviously has a different expression, lacks the visor, different paint job, and he also has this tank now.
8.167
em-000000.tar
EN_B00000_S00040_W000044
now it's time to compare them to a couple other helicopters. of course, starting just at the very beginning of all time, we have cathy copter and ron hover from the original cars movie.
11.189
em-000000.tar
EN_B00000_S00040_W000045
so obviously quite a bit different, but helicopters nonetheless.
4.074
em-000000.tar
EN_B00000_S00040_W000046
and roscoe, who i really do like both of these. i like roscoe a little bit more just cause i love that red and white color scheme. vasquez is kind of like a mexican helicopter. but yea, roscoe and kathy copter.
14.482
em-000000.tar
EN_B00000_S00040_W000047
are the perfect couple in my opinion. i mean, look at them. they're meant to be together. they're both rs10 helicopters. it's really a shame if they aren't a couple. i mean, come on.
9.083
em-000000.tar
EN_B00000_S00040_W000048
who should we go with next? how about hector vector, who i believe was the first helicopter to be released from the plains lime. he obviously appeared in the first movie, you know, again, a part of the mexican
11.257958
em-000000.tar
EN_B00000_S00040_W000049
very, very faint on some of these names for the planes, but thankfully it just popped into my head.
5.62
em-000000.tar
EN_B00000_S00040_W000050
that just popped right into my head. i was like, yes, that's falco. let's go.
4.448
em-000000.tar
EN_B00000_S00040_W000051
and last but not least, actually no, nevermind, not last, this is windlifter who's the biggest helicopter of them all. he's got a ton of blades up there and he's pretty much mainly meant for the fire.
14.126
em-000000.tar
EN_B00000_S00040_W000052
extinguisher retardant type thing where he puts out the fires obviously with his massive tank.
5.552
em-000000.tar
EN_B00000_S00040_W000053
so, uh, this one got re-released for moonmater with a black nose, right there.
6.468
em-000000.tar
EN_B00000_S00040_W000054
so i think we could just pretend that they're the same, because literally they are.
5.059
em-000000.tar
EN_B00000_S00040_W000055
and i think that's all the helicopters that have been released. i may have missed one because my memory is not phenomenal.
7.385
em-000000.tar
EN_B00000_S00040_W000056
and it actually is brief and humble, i just have so much i have to remember.
3.395
em-000000.tar
EN_B00000_S00040_W000057
so yea guys, thank you so much for watching this video. i do want to give a quick little announcement on the state of the disney docket channel. cause i've gotten a lot of questions lately like disney docket, what happened to all those videos you were going to record? you had so many on your video schedule in the about section of your channel and then you just removed all of them except for one.
19.987
em-000000.tar
EN_B00000_S00040_W000058
thing is, i recently started an internship, which i have mentioned before, and it literally is every day, monday, tuesday, wednesday, thursday, which i guess is not every day.
11.358
em-000000.tar
EN_B00000_S00040_W000059
and i also even work at the country club thursdays as well sometimes. so i mean, it's a ton of work. like i'm working a lot of hours a day. like i'm literally working around like.
8.574
em-000000.tar
EN_B00000_S00040_W000060
fourteen hours a day sometimes and it's kind of ridiculous. i'm kind of overworking myself, but
6.204
em-000000.tar
EN_B00000_S00040_W000061
need that cash, need that bmw i8. so anyways guys, i will not be producing quite as many videos, but you know, i started my internship today, came home, feeling good, decided to make a video. so the thing is i will still make quite a few videos, but i'm not going to commit myself to any schedules, like any strict schedules that will put pressure on myself.
20.668
em-000000.tar
EN_B00000_S00040_W000062
and just give me unnecessary stress. cause like if i schedule like, okay, i'm gonna do four videos a week like i used to.
6.108
em-000000.tar
EN_B00000_S00040_W000063
like that's just too much for me and i may be able to do those videos, but like it just puts so much pressure on me. i don't want you guys to hold me to those expectations that i likely will not be able to meet. so instead all the videos will just be like happy go lucky surprises like this one, which it was scheduled for today. but you know.
20.798
em-000000.tar
EN_B00000_S00040_W000064
i just wasn't sure if i was going to be able to do it, but i'm happy, you know, to still be out here and doing videos for you guys. and i will probably slow down a little bit as we get into august, because i need to pre-record a bunch of videos that will go out while i'm in college, because i will not be returning home from august 10th to like november 10th.
20.735
em-000000.tar
EN_B00000_S00040_W000065
so you guys can actually, you know, see some disney docket while i'm away. but yeah, very exciting times coming up. a lot of cool videos. you know, the one video i left scheduled, which i don't know if that will come out at that time, is the case k unboxing. but either way, you know, as soon as i receive that case, i will put the video out. so whether it be before that, whether it be after that, whatever it is,
22.246
em-000000.tar
EN_B00000_S00040_W000066
i will always do the case unboxings videos immediately just cause.
3.616
em-000000.tar
EN_B00000_S00040_W000067
you know, i want to get the new stuff out to you guys as soon as possible, but a lot of stuff i've been reviewing lately has been old stuff, but you know, that doesn't mean it's necessarily boring. so anyways guys, sorry for ramble. i just wanted to give you that quick little update. so i'll see you guys soon for another video. bye now.
16.893
em-000000.tar
EN_B00000_S00050_W000000
oh certainly, certainly very different from what anybody would do now or even in the last couple of hundred years.
5.365
em-000000.tar
EN_B00000_S00050_W000001
yeah, it's pretty flat, you can work with it.
3.14
em-000000.tar
EN_B00000_S00050_W000002
this is a side-axe, uhm, it's just ground on one edge, so it's flat on the other, so you can just...
8.506
em-000000.tar
EN_B00000_S00050_W000003
so there are some things you do need a saw for. so we'll just cut the pegs off to size.
6.52
em-000000.tar
EN_B00000_S00060_W000000
this episode of the rt podcast is brought to you by godaddy. it's gotime and godaddy is here to help you kick ass online. start your website today with $1.49.com. visit godaddy.com and enter promo code rooster149 to get your .com for just $1.49. some limitations apply. see website for details.
15.093
em-000000.tar
EN_B00000_S00070_W000000
there is a drug killing thousands of people in our communities. it can be hidden in all drugs. meth, heroin, pills. the tiniest amount of it could kill you. it's a substance called fentanyl. you may think it won't happen to you, but you are not the exception. don't be another face of fentanyl.
22.92
em-000000.tar
EN_B00000_S00070_W000001
learn more about the deadly dangers of fentanyl at facesoffentanyl.net
3.481958
em-000000.tar
EN_B00000_S00080_W000000
i said be sorry about sherby, you just looked the ball into your glove. choo choo.
4.635
em-000000.tar
EN_B00000_S00090_W000000
in a matter of time, the trial of the two men charged with your son's death will be concluded.
4.33
em-000000.tar
EN_B00000_S00100_W000000
right now, just getting your head right, you know, getting focused for the fight, making sure that, you know, the things we want to throw out there are on point. just breaking a sweat and making sure that weight's down for weigh-ins. but, uh, i mean the preparation's done.
11.765
em-000000.tar
EN_B00000_S00100_W000001
during camp, you kind of break your body down. this is kind of week where you get to rest a little bit, heal up, and just make sure everything's working right. we went through the hard work during camp and you're not going to get any better now. you're not going to get any faster. you're not going to get any stronger. so it's just that mental game from here on out. so, uh, i'm feeling good, man. my mind's right and i'm ready to go out there and win this thing.
19.49
em-000000.tar
EN_B00000_S00110_W000000
well i wouldn't say you needed work, i'd just say this one is just improved. you know, it's just, it's a different way, it's a new kind of version of doing it.
7.623
em-000000.tar
EN_B00000_S00110_W000001
thank you, dude. yea, i do slam poetry. no, but it's, it's, uh, it's a good, it's a good thing.
5.094
em-000000.tar
EN_B00000_S00110_W000002
but uhm, like i know i would go in there, i would find them, i would use them like instantly.
4.771
em-000000.tar
EN_B00000_S00110_W000003
i'd be like, yes, you know. i'd keep them in the inventory.
3.718
em-000000.tar
EN_B00000_S00110_W000004
i mean, i think i told you about that, matt. like that you could, when we were first starting to play, i was like, oh, you know, you can use these books and they'll, or the comics and they'll do.
9.151
em-000000.tar
EN_B00000_S00110_W000005
i think becoming so op is part of the charm of it though. i agree though, yeaah. cause you have to put so much time to get to that point. like my fallout 3 character is a monster. you know what i mean? but that took fucking time.
12.003
em-000000.tar
EN_B00000_S00110_W000006
the second you mentioned, that's why i pictured his picture, like his twitter picture. it's like the one in the suit, right? yup. i think it has a little hat on it, i don't know. he's gonna tweet when he hears this and he'll be like,
13.277
em-000000.tar
EN_B00000_S00110_W000007
yeah, right. part of the charm, i think, about fallout's music especially, and i'm talking about the radio music, not the ambient music, uh, is that the songs always have an underlying to what's actually going on in the world. so if you look at the trailer, you know, you have this song, it was like, atom bomb, baby. you know, you look at that, you know, stuff like that. so i think that's part of the
19.42
em-000000.tar
EN_B00000_S00110_W000008
the thing that we really love about it is that it has, like, an underlying, like, grim side to it.
6.62
em-000000.tar
EN_B00000_S00110_W000009
oh, that was so weird. i've listened to, i've listened to like the full thing of that. it's just weird. i don't know.
8.744
em-000000.tar
EN_B00000_S00110_W000010
i asked pete if he knew who i was last night.
3.905
em-000000.tar
EN_B00000_S00110_W000011
we really are. but that's what, that's our, you know what, okay, i'm gonna back up to a little bit something like deeper here.
6.197
em-000000.tar
EN_B00000_S00110_W000012
right, well, and see, that's just it. the way i look at it is, we were fine without season passes, you know, what, six years ago, five years ago, we didn't really have them that much.
9.338
em-000000.tar
EN_B00000_S00110_W000013
we can be fine without them again. just release the downloadable content as it's released. we'll pay for it. you know, i, i, i'm sorry, i just, i have a lot of issue paying $60 for a game.
10.34
em-000000.tar
EN_B00000_S00110_W000014
on top of spending another, say, how much is it? 40 or 20? 30. 30? okay.
5.365
em-000000.tar
EN_B00000_S00110_W000015
see, i don't think so though. i think it's gonna get worse. because, because you're right on the fact of pre-ordering kind of dwindling away in terms of the, uh, the digital downloads and stuff like that becoming more relevant. but the fact that, uhm,
15.195
em-000000.tar
End of preview. Expand in Data Studio

Raon-OpenTTS-Pool

RAON-OpenTTS

Homepage GitHub Hugging Face X License

Technical Report (Coming soon)

Raon-OpenTTS-Pool is a large-scale open English speech corpus for text-to-speech (TTS) training, constructed from 8 publicly available speech corpora and a set of web-sourced recordings. It is the training data behind RAON-OpenTTS, an open TTS model that performs on par with state-of-the-art closed-data systems.

  • 615K hours of speech audio
  • 239.7M speech segments
  • 11 source datasets aggregated into a unified format
  • All audio stored as 16 kHz mono Opus (64 kbps) in WebDataset tar shards

We restrict data sources to publicly available English speech datasets with more than 500 hours of audio. All speech segments are limited to 30 seconds or shorter to reduce alignment errors, multi-speaker content, and non-speech artifacts. Existing public datasets (LibriHeavy, Emilia, VoxPopuli, etc.) are included as-is without modification, with audio standardized to 16 kHz mono Opus 64 kbps for storage efficiency. The Raon-YouTube-Commons portion is reconstructed from YouTube-Commons through a dedicated preprocessing pipeline (see below).

With a model-based filtering pipeline applied to Raon-OpenTTS-Pool, we derive Raon-OpenTTS-Core, a curated high-quality subset of 510.1K hours and 194.5M segments.

For more details, see our paper: Raon-OpenTTS: Open Models and Data for Robust Text-to-Speech

Format

Each WebDataset tar shard contains pairs of files per sample:

{sample_key}.opus   # 16 kHz mono Opus 64 kbps audio
{sample_key}.json   # {"text": "...", "duration": 8.42, "source": "..."}

Note: The dataset viewer shows metadata only (sample_key, text, duration, shard_name). Audio is stored in WebDataset tar files — see Usage below to download and load audio.

Splits

Each dataset config has two metadata splits:

  • pool — all samples (sample_key, text, duration, shard_name)
  • core — quality-filtered subset (Raon-OpenTTS-Core), retaining ~85% of the data

Raon-OpenTTS-Core Filtering

Raon-OpenTTS-Core is constructed by applying three model-based quality filters and removing the bottom 15% of samples by combined score:

  1. WER-based: Transcribe each segment with Whisper-small ASR and compute WER against the existing text annotation. Samples with excessively high WER (> 0.35) indicate severe transcription mismatches.
  2. DNSMOS-based: Estimate perceptual speech quality using DNSMOS. Samples below 2.24 indicate strong background noise or distortion.
  3. VAD-based: Estimate speech activity ratio (SAR) using Silero VAD. Samples with SAR below 0.79 are dominated by silence, music, or non-speech audio.
  4. Combined: Compute an absolute rank for each segment along each criterion (DNSMOS, WER, SAR) and average the ranks into a single combined score. Segments falling below the 15th percentile are discarded.

This combined filtering achieves the best overall TTS performance across diverse evaluation benchmarks (see paper, Figure 3).

Available Datasets

Dataset Source Size (h) Avg. Dur. (s) Segments (M) Tars License DNSMOS WER SAR
Raon-YouTube-Commons YouTube-Commons 335k 8.5 141.70 1,017 CC BY 4.0 2.74 0.30 0.90
Emilia-YODAS2 Emilia 92k 9.2 35.97 287 CC BY-NC 4.0 2.82 0.19 0.90
Emilia Emilia 47k 9.3 18.14 145 CC BY 4.0 3.02 0.18 0.89
LibriHeavy LibriHeavy 42k 14.2 10.77 127 Public Domain 3.22 0.11 0.83
HiFiTTS HiFiTTS2 37k 10.1 13.09 109 CC BY 4.0 3.20 0.11 0.84
PeoplesSpeech-Dirty People's Speech 28k 14.2 5.48 63 CC BY 4.0 2.63 0.25 0.86
VoxPopuli VoxPopuli 17k 27.8 2.24 50 CC-0 2.82 0.36 0.83
PeoplesSpeech-Clean People's Speech 10k 1.50 18 CC BY 4.0
LibriTTS-R LibriTTS-R 552 5.6 0.35 2 CC BY 4.0 2.96 0.06 0.91
SPGISpeech2-Cut SPGISpeech 2.0 889 14.4 0.22 3 Kensho UA 2.72 0.08 0.90
Total 615k 9.2 239.7 1,821 2.83 0.24 0.89

Raon-YouTube-Commons

A substantial portion of Raon-OpenTTS-Pool (335K hours) is derived from YouTube-Commons. Since the original release provides only YouTube URLs with noisy or unreliable transcriptions, we reconstructed it into a high-quality speech-text dataset through the following pipeline:

  1. Audio collection: Download audio from YouTube URLs in the original dataset
  2. Source separation (UVR-MDX): Suppress background music and non-vocal components
  3. Speaker diarization (PyAnnote 3.1): Estimate speaker boundaries to ensure single-speaker segments
  4. Voice activity detection (Silero VAD): Segment continuous speech regions into clips of 3--30 seconds
  5. Automatic transcription (Whisper-large-v3): Transcribe each segment to obtain aligned speech-text pairs
  6. Standardization: Resample to 16 kHz mono, encode as 64 kbps Opus

The resulting dataset is released as Raon-YouTube-Commons in this repository.

Non-redistributable Datasets

Two additional datasets used in training cannot be included due to license restrictions. Users who have agreed to the license on HuggingFace can automatically download and convert them using prepare_nonredist_datasets.py:

Dataset Size (h) License Source
GigaSpeech 10k License agreement required speechcolab/gigaspeech
SPGISpeech 5k Non-commercial (Kensho) kensho/spgispeech

See Preparing Non-redistributable Datasets for instructions.


Usage

1. Metadata (pool / core split)

from datasets import load_dataset

# Core metadata for a single dataset
meta = load_dataset("KRAFTON/Raon-OpenTTS-Pool", "Raon-YouTube-Commons", split="core")
# Columns: sample_key, text, duration, shard_name
print(meta[0])

# All datasets combined
all_core = load_dataset("KRAFTON/Raon-OpenTTS-Pool", "all", split="core")

2. Audio (WebDataset, local tars)

Download tars first:

from huggingface_hub import snapshot_download

local_dir = snapshot_download("KRAFTON/Raon-OpenTTS-Pool", repo_type="dataset",
                               ignore_patterns=["*.parquet"])

Then load with WebDataset:

import webdataset as wds
import json, io, soundfile as sf

dataset = (
    wds.WebDataset(f"{local_dir}/LibriTTS-R/lr-{{000000..000001}}.tar")
    .to_tuple("opus", "json")
)
for opus_bytes, json_bytes in dataset:
    meta = json.loads(json_bytes)
    audio, sr = sf.read(io.BytesIO(opus_bytes))
    text = meta["text"]

3. Core-only training

The audio tars contain pool and core samples mixed. To train on core only, filter by sample_key:

import webdataset as wds
from datasets import load_dataset
import json, io, soundfile as sf

# Step 1: load core sample keys from metadata
core_keys = set(
    load_dataset("KRAFTON/Raon-OpenTTS-Pool", "LibriTTS-R", split="core")["sample_key"]
)

# Step 2: stream tars, skip non-core samples
dataset = (
    wds.WebDataset(f"{local_dir}/LibriTTS-R/lr-{{000000..000001}}.tar")
    .select(lambda s: s["__key__"] in core_keys)
    .to_tuple("opus", "json")
)
for opus_bytes, json_bytes in dataset:
    meta = json.loads(json_bytes)
    audio, sr = sf.read(io.BytesIO(opus_bytes))
    text = meta["text"]
    duration = meta["duration"]

Preparing Non-redistributable Datasets

The script prepare_nonredist_datasets.py automatically downloads and converts GigaSpeech and SPGISpeech into the same WebDataset tar + parquet format used by Raon-OpenTTS-Pool.

Prerequisites

  1. Accept the dataset license on each HuggingFace dataset page:

  2. Set your HuggingFace token (from an account that has accepted the licenses):

    export HF_TOKEN=hf_your_token_here
    
  3. Install dependencies:

    pip install "datasets<4.0" soundfile pyarrow numpy tqdm
    

    Note: datasets>=4.0 dropped soundfile audio decoding and requires torchcodec with system FFmpeg libraries. Use datasets<4.0 (e.g. datasets==3.5.0) to avoid this.

  4. ffmpeg must be in PATH.

GigaSpeech

# Download and convert xl subset from HuggingFace Hub
python prepare_nonredist_datasets.py gigaspeech \
    --output_dir ./GigaSpeech \
    --gigaspeech_subset xl \
    --num_workers 16

# Or from a local HF snapshot (no HF_TOKEN needed)
python prepare_nonredist_datasets.py gigaspeech \
    --source_dir /path/to/gigaspeech_local \
    --output_dir ./GigaSpeech \
    --gigaspeech_subset xl

Available subsets: xs (10h), s (250h), m (1000h), l (2500h), xl (10000h)

SPGISpeech

# Download and convert L subset from HuggingFace Hub
python prepare_nonredist_datasets.py spgispeech \
    --output_dir ./SPGISpeech \
    --spgispeech_subset L \
    --num_workers 16

# Or from a local HF snapshot (no HF_TOKEN needed)
python prepare_nonredist_datasets.py spgispeech \
    --source_dir /path/to/spgispeech_local \
    --output_dir ./SPGISpeech \
    --num_workers 16

Available subsets: L (full 5000h), M (1000h), S (~200h), dev, test

Output

<output_dir>/
  {prefix}-000000.tar     # WebDataset shard (~10 GB)
  {prefix}-000001.tar
  ...
  metadata_pool.parquet   # all samples
  metadata_core.parquet   # = pool (no quality filtering without --core_json)

By default metadata_core.parquet equals metadata_pool.parquet since quality filtering requires an internal index file. If you have pool_indices_filter_remove_15pct_combined.json from the Raon-OpenTTS maintainers, pass it with --core_json to generate a filtered core split.

Using with RAON-OpenTTS training

Once prepared, pass the output directory as a nonredist_dirs entry in the training config:

datasets:
  nonredist_dirs:
    - /path/to/GigaSpeech
    - /path/to/SPGISpeech

License

This repository contains data from multiple sources, each with its own license. Users must comply with the license of each individual sub-dataset they use.

Dataset License Commercial Use
Raon-YouTube-Commons CC BY 4.0 Yes
Emilia CC BY 4.0 Yes
Emilia-YODAS2 CC BY-NC 4.0 No
LibriHeavy Public Domain (LibriVox) Yes
HiFiTTS CC BY 4.0 Yes
PeoplesSpeech-Clean / Dirty CC BY 4.0 Yes
VoxPopuli CC-0 Yes
LibriTTS-R CC BY 4.0 Yes
SPGISpeech2-Cut Kensho User Agreement Non-commercial
GigaSpeech (non-redist) License agreement required See terms
SPGISpeech (non-redist) Kensho User Agreement Non-commercial
Metadata and dataset structure CC BY 4.0 Yes

Note: Emilia-YODAS2 and SPGISpeech2-Cut are licensed under non-commercial terms. If you require fully commercial-use data, exclude these sub-datasets via the configs parameter.

Citation

@article{raon2026opentts,
  title     = {Raon-OpenTTS: Open Models and Data for Robust Text-to-Speech},
  author    = {TBD},
  year      = {2026},
  url       = {https://github.com/krafton-ai/Raon-OpenTTS}
}

© 2026 KRAFTON

Downloads last month
55

Models trained or fine-tuned on KRAFTON/Raon-OpenTTS-Pool

Collection including KRAFTON/Raon-OpenTTS-Pool