[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / r / s / t / u / v / vg / vm / vmg / vr / vrpg / vst / w / wg] [i / ic] [r9k / s4s / vip / qa] [cm / hm / lgbt / y] [3 / aco / adv / an / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / pw / qst / sci / soc / sp / tg / toy / trv / tv / vp / vt / wsg / wsr / x / xs] [Settings] [Search] [Mobile] [Home]
Board
Settings Mobile Home
/wsg/ - Worksafe GIF

Name
Options
Comment
Verification
4chan Pass users can bypass this verification. [Learn More] [Login]
File
  • Please read the Rules and FAQ before posting.
  • Supported file types are: GIF, WEBM

08/21/20New boards added: /vrpg/, /vmg/, /vst/ and /vm/
05/04/17New trial board added: /bant/ - International/Random
10/04/16New board for 4chan Pass users: /vip/ - Very Important Posts
[Hide] [Show All]


[Advertise on 4chan]


File: agent-47-funkytown.webm (4.4 MB, 1268x720)
4.4 MB
4.4 MB WEBM
Previous thread >>5493415
Due to high interest, there is now a dedicated suno thread >>5506204

Post anything AI generated. Song covers, animations, etc.
OC encouraged, but not required.
This thread focuses on audio and video with an audio component.
Let me know if you have more links to add. This thread is a work in progress.

> Voice-to-Voice
RVC walkthrough (somewhat outdated, collab is dead): https://docs.google.com/document/d/13_l1bd1Osgz7qlAZn-zhklCbHpVRk6bYOuAuB78qmsE/edit
Models, mega links, and mirrors: https://docs.google.com/spreadsheets/d/1tAUaQrEHYgRsm1Lvrnj14HFHDwJWl0Bd9x0QePewNco/edit#gid=0
https://github.com/Mangio621/Mangio-RVC-Fork
https://github.com/Vali-98/XTTS-RVC-UI
https://github.com/voicepaw/so-vits-svc-fork

> Text-To-Speech
https://github.com/collabora/WhisperSpeech
https://github.com/myshell-ai/OpenVoice
https://github.com/yl4579/StyleTTS2
https://github.com/BoltzmannEntropy/xtts2-ui
https://github.com/daswer123/xtts-webui (Warning: Windows version uses prebuilt binaries that anons haven't verified. Use at your own discretion)

> Music
> Online / Freemium
https://www.suno.ai/
https://www.udio.com/
https://www.stableaudio.com/
> Offline / FOSS
https://github.com/facebookresearch/audiocraft
https://rentry.org/AudioCraftRemix

> Vocal Cleanup
https://github.com/Anjok07/ultimatevocalremovergui
https://github.com/resemble-ai/resemble-enhance

> Related boards
>>>/g/sdg
>>>/g/lmg
>>>/g/aicg
>>>/h/hdg
>>>/e/edg
>>>/d/ddg
>>>/b/degen
>>>/vt/vtai
>>>/aco/asdg
>>>/aco/csdg
>>>/trash/sdg
>>
File: mint-bustin.webm (4.13 MB, 1066x720)
4.13 MB
4.13 MB WEBM
Extra reminder that there is now a dedicated thread for suno.ai due to the high amount of suno posts relative to other content.
>>
File: lotr-short-intro.webm (4.38 MB, 720x576)
4.38 MB
4.38 MB WEBM
>>
can anyone with an rtx 4090 use subtitle edit's ai subs (engine:purfview's faster-whisper) with a jav downloaded from https://sukebei.nyaa.si/ (must be at least more than 2 hours and 30 minutes like ABF-094) and tell me how long the eta was along with the actual time it took and with the code of the jav you used also with screenshot for confirmation
>>
>>5504578
What can possibly top this?
>>
>>5515280
This.
>>5506729
>>
>>5515281
it's fine but not even close to as funny
>>
File: the-story.webm (2.62 MB, 1280x720)
2.62 MB
2.62 MB WEBM
Anyone know where that riff at 0:10 comes from? I swear I've heard it somewhere before
>>
>>5515283
>I swear I've heard it somewhere before
Sandstorm by Darude.
>>
>>5515198
I'm on it. Wait for me here.
>>
>>5515457
hello?
>>
File: House Party Bash.webm (2.39 MB, 720x720)
2.39 MB
2.39 MB WEBM
>>5515011
I'm having a lot of fun with the software
>>
File: Canciron del Pirata.webm (2.4 MB, 552x552)
2.4 MB
2.4 MB WEBM
>>
File: 1704787632783944.webm (3.95 MB, 720x480)
3.95 MB
3.95 MB WEBM
>>
ym4dm
>>
File: BATMAN.webm (3.1 MB, 854x480)
3.1 MB
3.1 MB WEBM
>>
File: 1684424900620419.webm (5.41 MB, 320x180)
5.41 MB
5.41 MB WEBM
>>
>>
File: Megumin - leekspin.webm (2.83 MB, 320x320)
2.83 MB
2.83 MB WEBM
>>
>>
File: stukkie wukki.webm (636 KB, 586x302)
636 KB
636 KB WEBM
>>
File: 1685266785118026.webm (2.26 MB, 408x720)
2.26 MB
2.26 MB WEBM
>>
>>5516725
Oh , noooooooooooooooooooooooo
Poor people steal from the rich.
Noooooooooooooooooo
I am amrimutt capitalist lover. I always side with rich. Only they can steal and exploit the poor. The poor should never steal from the rich.
All my problems are because of the jews and because of the niggers.
>>
File: 1688609170664529.webm (666 KB, 512x768)
666 KB
666 KB WEBM
>>
File: geralt.webm (836 KB, 692x414)
836 KB
836 KB WEBM
>>
File: geralt2.webm (262 KB, 830x891)
262 KB
262 KB WEBM
>>
File: devil.webm (3.82 MB, 800x384)
3.82 MB
3.82 MB WEBM
>>
File: triss.webm (1.43 MB, 1280x640)
1.43 MB
1.43 MB WEBM
>>
File: oldfag brooks.webm (5.81 MB, 480x360)
5.81 MB
5.81 MB WEBM
Interest for AI voice funnies has really tanked huh?
>>
File: tpd-now.webm (1.1 MB, 480x480)
1.1 MB
1.1 MB WEBM
>>5515011
Wanted to polish this up for my YouTube channel, but decided it wasn't worth it. Here you go, guys.
>>
>>5519545
i can understand that since the pasta got stale in 2022
>>
>>5519552
Eh... I feel like it could've worked if I came up with some funnier things to do to postmodernists. I just kinda ran out of ideas lol
>>
File: rock-with-yuu.webm (4.59 MB, 1064x720)
4.59 MB
4.59 MB WEBM
Speaking of youtube, anyone else upload covers there? I just ran into a weird issue with the content detection system. I did a cover album and it seemed to allow quite a few songs at first, then when I went to make it public it changed its mind and decided all songs were blocked in all regions. I'm not sure what happened but it seems like it decided to block them all when I added descriptions.
Anyone else experience this weirdness? It's like impossible to know what will and won't be blocked now. One of them was this, and historically the Michael Jackson catalogue has been fine. You think they're specifically blocking AI covers now?
>>
File: 1707751697455051.webm (2.9 MB, 600x600)
2.9 MB
2.9 MB WEBM
>>
File: 1675825450603696.webm (637 KB, 1080x1920)
637 KB
637 KB WEBM
>>
File: 1677061457265442.webm (4.66 MB, 960x720)
4.66 MB
4.66 MB WEBM
>>
>>5519554
people keep making that pasta with ai, if i've heard it once i've heard it 100 times so it's an instant close after 1 second. i can't imagine any anon sitting there listening to the whole thing, eager to find out how that particular voice will pronounce the same words you've listened to so many times before.
>>
File: lynchian.webm (1.81 MB, 960x540)
1.81 MB
1.81 MB WEBM
>>
File: if you insist.webm (4.07 MB, 720x720)
4.07 MB
4.07 MB WEBM
>>
File: lamp.webm (1.93 MB, 800x480)
1.93 MB
1.93 MB WEBM
>>
>>5515283
what a time to be alive
>>
>>5519600
Well... I mean I didn't just do the usual list of deaths you hear in a "kill / behead / roundhouse kick" pasta. Sure I did the first three (you have to do the first three), but after that I tried to get creative.
Here's the full pasta I wrote for him:
> Kill postmodernists.
> Behead postmodernists.
> Roundhouse-kick a postmodernist into the concrete.
> Send a postmodernist into space without oxygen.
> Gaslight postmodernists into thinking they aren’t smart enough to understand the artistic reasons why you’re killing them with actual poison gas.
> Light a postmodernist on fire.
> Push postmodernists into industrial lathes.
> Strap a postmodernist’s balls to a surgical table under a hydraulic press, and say you’ll free him if he can define “crush” before the machine finishes a cycle.
> Use false rape accusations to ruin a postmodernist’s life.
> Grind up a postmodernist in escalator machinery.
> Trap postmodernists in an elevator before cutting the power and severing the cables.
> Hang postmodernists from every lamppost, up and down Madison avenue.
> Throw postmodernists out of helicopters.
> Suck postmodernists up into jet engines.
> Hurl postmodernists into active volcanoes.
> Cancel postmodernists on Twitter.
> Grind up a postmodernist’s baby in a woodchipper, point the discharge chute at a canvas, and tell the parents the result is a long-lost Pollock before demanding they defend its artistic merits.

> Total Postmodernist Death Now.
>>
>>5516742
Pretty good but annoying how you can tell when the AI and normal voices cut in and out
>>
>>5519579
Can you make an Odysee or Rumble channel for your blocked stuff Chameleon anon? It should be better, especially for music stuff.
>>
>>5520074
Yeah that's probably the way to go. I keep telling myself I'll make an Odyssee account and I think I'll work on that this weekend.
>>
>>5520074
>>5520123
i thought odysee was about to be shut down because they had financial problems? have they been sold or is hanging on by a thread?
>>
File: yuu-make-loving-fun.webm (5.94 MB, 720x414)
5.94 MB
5.94 MB WEBM
>>5520074
I feel like there was little point in putting this one on Odysee since it's a joke based on an already obscure vtuber, but here it is.
https://odysee.com/@ChameleonAi:1/yuu-cover-album:e
>>
File: kek.webm (5.99 MB, 1024x768)
5.99 MB
5.99 MB WEBM
>>5518697
>>
File: dunkin.webm (2.81 MB, 854x480)
2.81 MB
2.81 MB WEBM
>>
File: 1711365530999492.webm (1.67 MB, 686x384)
1.67 MB
1.67 MB WEBM
>>
File: sonic.exe.webm (1.23 MB, 870x870)
1.23 MB
1.23 MB WEBM
>>5521339
i wanted to post the one where sonic and shadow runs next to each other talking, but lost it. anyone have it? not sure if it was ai, i think not.
>>
File: soldic.webm (5.18 MB, 500x500)
5.18 MB
5.18 MB WEBM
>>
StyleTTS2 output file put through RVC. Only the RVC model was fine tuned. Karolina Zebrowska audio I gathered was denoised with Resemble Enhance and de-reverbed with Izotope RX 10.
>>
for the requester in the waifuy /co/ thread. it turned out pretty good imo.
https://vocaroo.com/1fzbtfJvVoIB

For the huntress wizard requestor (if you're around), it may take a little bit longer than tonight because I need to find her voice clips and train a model for her, but I haven't forgotten.
>>
>>5521731
damn I really butchered that first sentence
>>
>>5521678
Sounds like you trained a decent RVC model. The TTS sounds like TTS but that ain't bad considering it's all local.
Isn't Izotope paid software? Ultimate Vocal Remover has de-reverb and de-noise models, in case anyone is interested in the possibility of a fully FOSS solution. UVR is pretty versatile and not just for separating music.
>>
https://vocaroo.com/1hKUqFxegPOr
>>
is whisper ai good for javs?
>>
>>5521731
https://vocaroo.com/1eOMEWUsQFBf

Here's the Nya Nya Huntress Wizard.
the model turned out great lol, thank god she has a standup special so I could get so many clips
>>
https://vocaroo.com/1iONv5xfSD7A
>>
File: red dead redemption.webm (5.91 MB, 498x280)
5.91 MB
5.91 MB WEBM
>>
>>5515011
I wish I could still enjoy hearing funkytown.
>>
>>5523327
Hey Anon, why the long face?
>>
>>5515011
Ice Ice Matrix AI edit
https://www.youtube.com/watch?v=gnEIeVWLtbU&pp=ygUKYXVyYWxuYXV0cw%3D%3D
>>
who can convert this to a webm?

files.catbox.moe/zsgi5k.mp4
>>
I stumbled upon some anons talking about the spiked forest of nuclear waste sites, so I made an AI music video of it.
The lyrics are loosely based on this:
https://en.wikipedia.org/wiki/Long-term_nuclear_waste_warning_messages
Also, if anyone could give me pointers on how to make higher resolution webms with lower file size, that would be helpful.
>>
>>5522243
I haven't tested it on them, but whisper tends to have trouble with not clearly pronounced text.
JAVs have both massive noise and extremely irregular pronunciation.
>>
>>5523866
ffmpeg. If you have any questions, consult the man page.
>>
>>5523890
There's only so much you can do. Use vp9 and 2-pass encoding, reduce the frame rate, if audio is less important, reduce audio bitrate.
For that webm since it's a slide show you can get away with like 12fps.
>>
File: ghost.webm (4.79 MB, 640x480)
4.79 MB
4.79 MB WEBM
>>
>>5524504
Bit of a banger you've got there, mate.
>>
>>5524504
yeah dude, that's fucking good.
>>
>>5524504
holy shit, its actually good
>>
>>5524504
really makes me think of Within Temptation
>>
>>5524029
I'm just jealous of some of the webms in the anime threads. Some of them are just pure quality.
>vp9
Check VP9/Opus, right? That's already checked by default in Webms for lazys.
>2-pass encoding
Not sure where to find that in WfL.
>12fps
I wanted to set it to 15 instead of 30, but the program just stuttered at the start and refused to convert. Maybe I'll check the video editing program for lower frame rates next time.
>>
>>5518706
lol, nice
>>
>>
>>5521531
honestly one of the most banger AI covers
>>
>>5517608
Based, what was the prompt for this?
>>
>>5523890
meanwhile
>waltzes into the pyramids and takes the treasures
>>
>>5525062
dutch, sea shanty, trains - style
the lyrics are just whats on the sign
>>
>>5524601
If you can't specify 2-pass, use something else like Handbrake or ffmpeg command line.
>>
Did a couple of these as a joke cause a friend of mine plays Fallout 76. Turns out HL Scientist works well for older songs.
>>
>>5525200
>>
This was wonderful anon. Thank you for making this for me.
I loved it so much, I drew her singing it, just for you.
Something is preventing me from posting the picture so please take this catbox instead.
https://files.catbox.moe/vvqyny.png
>>
>>5523041
I can't listen to the original anymore. It has to be the WoW Parody of it.
>Grind baby, grind baby, but not the leveling kind.
>>
File: HLScienWorldOnFire.webm (2.63 MB, 320x240)
2.63 MB
2.63 MB WEBM
>>5525203
Did this one too but it's not as good as the other two. (Also he kinda has a stroke near the end for a bit.)
>>
>>5524029
i'll add a trick, blurring the video a bit could make it compress better (this is for very low bitrate visuals since otherwise you'll see the blur).
also lowering framerate too much doesn't work all that well for videos that doesn't change much, codecs usually understands that it is still images and won't repeat data. what you want to increase is the max time between keyframes since data is forced to repeat then.

>>5524601
anime/cartoon tends to compress well with the clean colors everywhere that stays the same between frames, don't expect to get that quality
>>
File: Southin Park.webm (3.99 MB, 480x480)
3.99 MB
3.99 MB WEBM
>>5524633
i think this south park cover is my favorite, the voices fits well and sounds right
>>
File: out of touch.webm (1.56 MB, 728x720)
1.56 MB
1.56 MB WEBM
>>5525200
>>5525203
you're right, suitable voice for those songs
>>
>>5525205
BLESS Anon! I love it!

And you probably can't post it because it's not animated
>>
>>5525320
If I knew how to animate, I would have her sing for you. Ah well. Catbox is just as good. I'm glad you like her. I enjoy drawing Marcy.
>>
>>5525329
If you have any ideas for Marceline, please let me know.
>>
>>5525329
you're a king anon. thanks for the art!

>>5525341
another anon making marceline covers, my collection grows doubly
>>
>>5525346
Hmm. I need to hear some more Marcelines. Post some random one and I'll take an idea.
>>
File: 1349789609146.gif (134 KB, 250x120)
134 KB
134 KB GIF
>>5525363
https://vocaroo.com/1h8iKnru5et3
https://vocaroo.com/1lKswf7BCtqw

are a couple of my favorites I've made of her. She's got a great voice for songs without too much vocal fry
>>
>>5525367
Mhmmm. I think I may have just the right idea.
>>
[1 / 3]
>>
File: Silly Music.webm (2.21 MB, 512x768)
2.21 MB
2.21 MB WEBM
>>5525693

[2 / 3]
>>
>>5525694

[3 / 3]
>>
File: preacher.webm (4.43 MB, 480x270)
4.43 MB
4.43 MB WEBM
>>
>>5525693
Can you read?
Dedicated Suno / Udio thread
>>5518247
>>
>>5525205
>Something is preventing me from posting the picture
You can't post pngs on /wsg/, either convert to .gif or post catbox links.
>>
>>5525964
better thread
>>
File: Tiny-Tim-Hl2-Zombie.webm (1.96 MB, 300x374)
1.96 MB
1.96 MB WEBM
fresh oc
>>
>>
>>5525212
Another one.
>>
File: txf-pepe.webm (4.81 MB, 720x480)
4.81 MB
4.81 MB WEBM
OC
>>
>>5526671
saved
>>
>>5526278
pls don't ruin this thread just because your music schlop becomes a drop in a bucket in the proper suno thread
>>
>>5527158
lol very good
>>
File: SFS.webm (5.69 MB, 720x480)
5.69 MB
5.69 MB WEBM
>>5515011
>>
File: six_million.webm (2.2 MB, 512x768)
2.2 MB
2.2 MB WEBM
An A.I. classic.
>>
is there a software somewhere that would allow me to remove a specific song from another audio track? (i don't mean vocal isolation, i mean actually removing one track from another track)

tried doing it manually with audacity but the result was pretty scuffed, but the only AI-supported applications i've found so far have been for vocal isolation
>>
>>5528072
Unclear what it is you want to split out, but have you looked at all the different Ultimate Vocal Remover models? There are MDX-NET models that split bass, drums, noise, and other stuff.
If you're looking at something like a movie where you'd have to split out sound effects from the backing track, I'm not sure if there's a good model for that, but in general UVR would be the right tool if someone trained a model for it.
>>
>>5528232
i tried a couple of tools, including UVR, to split the track but it didnt work how i hoped. it is movie footage, where there's a song playing (which has its own vocals) and also characters talking and also sound effects, and i'm trying to get just the characters talking + sound effects, or at least just the characters talking without the song's vocals. sadly most models seem to split the singer and the dialogue into the same track, which is why i'm hoping to find a tool that will remove an entire song that i give it because i have the song in question, separate from the movie, but the stuff i've tried on audacity hasnt worked out too well.
>>
>>5528249
Try UVR MDX-Net Main on the first pass, then Karaone 2 (or VR Karaoke) on the vocal track. That will try to separate overlapping vocals. Getting the sound effects out of the instrumental track is a different problem. I don't know of a model that can do this. I don't know how it would know what is "song" and what is "sound effect". I'm not aware of any tools that can take in a song as truth data and separate anything that's not that.
>>
File: DeepFaceLive.webm (5.81 MB, 720x404)
5.81 MB
5.81 MB WEBM
>>
>>5528249
>>5528523
I'd do VR Arch - UVR-BVE-4B_SN-44100-1, it's meant for separating two or more voices singing at the same time.
>>
>>5528709
I should clarify that I'd use MDX23C-InstVoc HQ first, then BVE on the isolated vocal track. Not sure about the sound effects, though. You could put the instrumental into Audacity and just cut out all of the music manually.
>>
>>5528523
>>5528709
sadly neither of these worked-- both sets of vocals ended up on the main track. i figured there might be some sort of tool which could read a specific noise (e.g. a song) and remove just that noise from an audio track, which would avoid the guesswork involved in pure AI tools, but it seems like nobody's made a tool like that. thank you for the suggestions anyways
>>
>>5528898
You would think that would be a thing, but someone smarter than me can probably explain why there's no tool that can separate out the difference between input audio vs the same audio comped into something else.
>>
>>5515011
why is there /g/aicg in here but not /vg/aids? Aren't they one of the first generative ai general, or are we gatekeeping?
>>
>>5524623
damn I feel like grandpa over here with rvc, i've been out of it for a few months and you guys have posted some pretty good shit, I love this one. >>5524504
>>
>>5529349
Of course not. I just inherited the list from another that didn't have it, I'll put it in.
>>
this thread will live
https://vocaroo.com/10lDvNbtFV86
>>
>>5528524
this needs to be stickied on the front page of twitch to constantly remind everybody
>>
File: ララ.webm (3.9 MB, 852x478)
3.9 MB
3.9 MB WEBM
>>5515198
>needing subs for JAV
ngmi
>>
Love the Witcher posts
>>
>>5515283
Holy fucking kino.
Entire jewish media on suicide watch.
>>
>>5519493
I spent several days and attempts trying to make an Emma Watson model. It never worked and then using them for text to speech was ridiculously slow making it impossible to use for books.
>>
>>5530531
How did you source the audio? I find that I get best results when the audio levels are fairly consistent across all clips. They can be from multiple sources as long as the quality and levels are fairly consistent across them all. Best case scenario would be if she read an audio book or did a podcast. Gathering a bunch of short clips from random places is not even worth trying, it'll sound like crap.
>>
>>5526671
based
>>
>>5526671
kek and saved
>>
File: announcement.webm (2.17 MB, 300x365)
2.17 MB
2.17 MB WEBM
posting a classic
>>
>>
>>5531626
god these are so cursed. 10/10
>>
This is not done yet but i probably won't finish it anyway
https://voca.ro/16KjKsMvnfjr



[Advertise on 4chan]

Delete Post: [File Only] Style:
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.