Last Thread: >> No.6097998https://suno.comhttps://udio.com>(since Udio disabled all downloads you need now either a browser extension, the yt-dlp script or you could get the audio file from your browsers developer tools/menu)https://www.producer.aihttps://sonauto.aihttps://www.choruz.aihttps://www.mureka.aihttps://elevenlabs.io/app/musichttps://www.minimax.io/audio/musichttps://huggingface.co/spaces/ACE-Step/ACE-Step
you have to post 3
>>6121631previous thread: >>>6097998
all of this is abysmal dogshit and it will remain abysmal dogshit for eternity
>>6121835don't know the fun in making abysmal dogshit or doesn't know how to make abysmal dogshit itself. you must not be a fun person to be around.
>>6121835Counterpoint: it's fun.
>>6121835Here you go.
>>6121631
>>6122384this slaps yo
>>6122384>>6122470saved, very nice
baby andy W
>>6121631space song demonstration of latest acestep xl turbo model. no loras, default model capabilities. saturated prompt = sound too clean.first release versions of acestep models from previous thread, sft variant of ace step v1.5;200 epochs steps suick lora>>6101550500 epochs, instrument training>>6103971700 epochs training>>6111070
>>6126328>suick**quick
>>6121631I made an AI band.https://www.youtube.com/watch?v=3jGIo68cpHA
>>6127920>I madenope.
>>6128746> still upset over the semantics of the word "made" in current yearshiggy diggy
https://files.catbox.moe/9w4ofy.mp4
Hello anons, I have a search request. Not sure if it's the right place but, does anyone else have fakemusicbr's original version of paradise city? It's this one on instgram, but it's only 3 minutes long. And i know for a fact they once uploaded a full version on youtubelinki: https://www.instagram.com /reel/DPBfRaAjeCy/Thanks in advance.
>>6128761nta but you didn't make anything though.
>>6128748@grok is this real?!
>>6127657great song
>>6126328xl sft version lora simple run
I find this one funny.
space song with lora on acestep xl sft
need tomboy gf
>>6122384baste
Short and catchy, lyrics made by hand.
>>6124497>>6124498>>6124499Kino & subscribed
>>6134629Thank you anon I appreciate it!
>>6121631Coming back after an exceptionally long IP range ban, and with a new SSD, I wanna ask if anyone has mega links or such to the-X-mas songs like Rammstein or Marcelinne-The heavy metal Sword and Sorcery faggotry which was popular at one time-the Vegeta meme collectionthanks in advance for giving a fuck
>>6135792I'm not sure if you meant these, but I posted some Rammstein Christmas songs from UDIO a while ago. But I didn't made them, I just found them there.
>>6135792>>6136268
>>6134629hood resurrection is based and you really got something good on that one. if you intend to make more theres plenty more to work with like cotton fields and whatever comes to mind for "Jigaboo Jammin"
>>6136268>>6136269>>6136534Wow this is dogshit.
>>6130510Truly a master piece.
>>6136268good shit
wuts the current year method to replace lyrics on audio song cover?the google colab from 2023 mo longer works
>>6138453
>>6128748best
https://litter.catbox.moe/4g1tjxhvqu4tborv.mp4Good: Gemini allows for genning longer songs now. Bad: they're on the bigger side, and somehow converting to WEBM makes them even bigger
>>6136268>>6136269oh nice I found these again, tyvm :D
https://litter.catbox.moe/054wykkjmn6ogq1f.mp4
>>6128748This goes harder than it has any right to.
www.producer.ai is now flowmusic.apphttps://www.flowmusic.app/playlist/5bf5401f-ef0f-495c-8035-00c090d0cf0c
how do you do shit like this though? I want the same song but in another genre, grok is retarded and keeps telling me to RVChttps://youtu.be/nSD0eF9JtDM
Fug does anyone have this one? I lost it https://desuarchive.org/wsg/thread/5652124/#5657466
>>6133409what model did you use for that? lipsync thing is impressive.not a bad song too.+ any of you learnt some acestep training tricks so far?
>>6143637Model is LTX2.3 1.1 Distilled. So I cut the audio into 20 sec sections and then just used a image for it to gen. LTX can do more than 20 secs but needs to gen an additional window which it then joins together. Far better just to stick to 20 secs (or 481 frames). Then I stitched it together in Shotcut which is an excellent free video editing program.
>>6143833>fat bitch riding a dildoFuck off to /gif/ you dumb cunt.
>>6143833actual porn warning, glad my gf is asleep
>>6143685ty.and lyrics are done via simple prompt where place whatever it needs to say?rough processing time and vram requirements plz, am vramlet.
>>6144127Well the music (so audio) is just Sudio, no need to put anything in the prompt other than what you want the gen to do visual wise and perhaps direct it to lip sync to the the audio. So the audio can be anything, you could for example use actual songs, copyright. For LTX2 you need at least 12gb vram, 16gb would be better. The processing time depnds on the resolution used for generation. It is fast compared to Wan especially as Wan you are only doing 5,6 or 8 sec clips. So the time to gen would be around 4 mins for a 20 sec clip at 540p, maybe 6 mins at 720p and nearer 20 mins for 20 sec clip at 1080p. You can upscale say the 540p and 720p but the genning done at a higher resolution is always better, interesting that you can use the exact settings, seed number and whilst doing that at the same resolution would result in an exact duplicate of the gen, doing so at a different resolution would result in a different gen. Taking int everything account it would not take that long to do even a 5-6 min song, I did that at 540p and upscaled using the upscaler inbuilt in the UI - so post processing. If i did it again I might do it at 720p and not bother with upscaling. All of this is on a 5060ti 16gb with 64gb of ram, so times will be different with lower and higher cards of course. Running out of ram either Vram or Ram is always an issue that can occur, however the models do use ways to offset that by dumping things into memory if VRAM isn't there and even uses paging and uses the hard drives. But on occasions it may say there just isn't enoug hram, i have had no problems with even doing 20sec clips and even longer at 1080p aside from doing a lot of gens one after the other and the PC simply reset. You just have to understand there are limits, if you have a higher card with the VRAM able to take anything then there are no real limits with these models as they are designed down to the lower cards.
https://vocaroo.com/1448AXu5IfVE
>>6144208info = turbo detailedty anon :)
https://files.catbox.moe/57kmo0.mp4
soft rock ballad
>>6146247abysmal dogshit
>>6146249>quickest (You) i've ever gottenlmao
https://www.dailymotion.com/video/xa9j9jo>https://www.dailymotion.com/video/xa9j9johttps://www.dailymotion.com/video/xa9j9jo>https://www.dailymotion.com/video/xa9j9johttps://www.dailymotion.com/video/xa9j9jo>https://www.dailymotion.com/video/xa9j9johttps://www.dailymotion.com/video/xa9j9jo>https://www.dailymotion.com/video/xa9j9jo
Not by me
Does anyone know anyway to get around Suno's copyright detection? I'm trying to do another cover of a song like this, except instead of "Johnny Cope", it's "Janny Cope" and it's going to be about Jannies coping.https://www.youtube.com/watch?v=EeIufFgQf44I've tried changing the speed, pitch, even a vocal remover but nothing works.
>>6149009Try using the wah-wah effect in Audacity. I could bypass most of the copyright bullshit with that.
>>6149009>>6149038Did some digging and turns out sonauto doesn't give a shit about copyright.It seems to stick to the style of the original audio better too.
we getting there ngl (both Suno AI song and Kling AI video lip-sync option)
>>6149046>>6149009ace step has a cover replace lyrics featurehttps://huggingface.co/spaces/ACE-Step/ACE-Stepany1 got it working?
>>6149009I did find a way via making a copy of the song that was copyright (so a song that was allowed to have remixes for a compitition and then didn't allow anyone to save the song). So you edit the song in the editor (you need a paid account tho for that) and simply selected all the song and crop it and save it as a new song. That song then is able to do anything any other song could do with the same lyrics intact so that would also mean any lyrics changed would also be allowed. another way would be to simply paste the original lyrics into the lyric box and then use the AI enhance function and ask it to change the lyrics to avoid any/all copyright but keep the same mood, structure etc. Then once you get the new lyrics simply edit them closer to the original and see how close you can get. To be honest I find it more creative to simply rewrite original lyrics and that enhance prompt function does a good job now - so much that using chatGPT might be redundant if using AI to write or rewrite lyrics
>>6149711did u try ace step?
>>6149749sorry, I was another anon replying making a suggestion - should have made that clear as I would just rewrite songs using AI. However another way would be to use voic cloning. Take section (or all of the audio) of the original audio and isolate the vocals into a stem using the audio tools in Sudio or use Audacity with the AI addons (you download them once audacity is installed) and once you have the vocals isolated you can use that as a reference to then use in a text to speech program. And you type in what is required and hopefully it sings as per the original vocal used as a reference. Then it is a case of stitching that audio into the sections you want - as you are only messing with the vocal track you won't be messing up the instrumental ones. How effective and seamless that would result in depends on how long you mess around on it, you can get audio levels the same with matching the gain so the stitched levels don't ramp up in volume or below the originals volume. I would imagine once instruments are present then the difference is less and less noticeable. the irony of course is that suno audio editor (not eve nthe sstudio one that costs more money to access) can replace lyrics for just a few credits, but if the copyright thing is noticed it won't allow it. I wonder tho if a small section of copyright music could be uploaded and edited and get by the censorship? It would make it very easy then, just upload the various sections and use the AI in the editor to replace with what you want.
>>6121631I hate this ugly thumbnail so much.
>>6149760there's also stuff like rocaloid, heartMula, vits svcand even converting the vocals to midi and then have ai gen the song with same melody, voice but new lyrics
>>6149009>>6149749I tried AceStep via Pinokio (the UI platform I use for AI) and it seems to work well and very fast (33 sec for this 1 minute sample). Now this was a first attempt and I tried to upload audio from that original but got 5 mins of audio of just the starting hornpipe), So genned just using a prompt and the lyrics. Maybe not the best example to see how well it fares as it is scottish and difficult to understand certain words. But will try with other copyright songs to see how well it goes and if I can get the audio in to work.https://voca.ro/1clP3CBPt0MZ
>>6150669Another attempt with the full 4 mins available from prompt only, get's noisy and distorted and song really finishes at 3 mins anyway.https://voca.ro/1iT0nAu9WeSH
>>6150669supposedly you may have more control and better results via cli which ace step also allowsAudio Continuation - Continue from existing audio with new lyrics, but same voice and etcAudio Repainting - Modify specific parts of audio
>>6150777I just looked and there is AceStep XL Turbo model with more features including a cover mode, transfer audio timbre (of the audio upload), BPM slider, temperature slider, keyscale, time signature, language. All easy on the Gradio UI in Pinokio as any models can simply be picked and it will download it before generating (only need to download once of course). Amusing tho because I had no interest in using anything other than Sudio and also if that other anon had asked how to do it (the original start of the conversation on this) and yet another anon had not mentioned Ace Step I would not have bothered. Two genned songs of "Janny Cope" (whether it sounds like Janny or Johnny is up for debate) using Ace Step XL Turboaudio in using the original with the start hornpipe not includedhttps://voca.ro/1jNGPisYd1mpprompt onlyhttps://voca.ro/12FmIfx02jp2The prompt only one is far better quality, doesn't sound as AI or as if done using MIDI instruments. However that might well be down to how it is tweaked and also if the prompt neds to compliment the audio. In any case it certainly has lot's of potential and can be used with any copyright lyrics and any lyrics.
Van (Morrison) Halen: Jumphttps://voca.ro/1gIwq4dwGtUE
>>6150852that's pretty cool. gonna have to try that out but aiming to do it all via CLI
Olivia Newton John: Physicalhttps://voca.ro/1d89nDGFCIBahttps://voca.ro/1344urRKAN1Ghttps://voca.ro/16NSwFt0LzSFhttps://voca.ro/1i5yWJOX1R2A
>>6149009ask the anon that made this >>6122384
Nirvana: Lithiumhttps://voca.ro/1kkF729ph6vShttps://voca.ro/15OAfEhaI5u0
Michael Jackson: Beat Ithttps://voca.ro/15fuqPSEtKdK
Rick Astley: Never Gonna Give You Uphttps://voca.ro/1mxXetH5MSohhttps://voca.ro/1LbdJNxIPkabhttps://voca.ro/1oKaIJ3NXszAhttps://voca.ro/13Rz0gFB36Mdhttps://voca.ro/13EQoAqhWxah
Not a single good post.
Ein Volk, ein Reich, ein Führerhttps://voca.ro/1oN6HhKntyjYhttps://voca.ro/16xZDUzzG9Mnhttps://voca.ro/1lrQChVEmiXRhttps://voca.ro/15HGa4SrdFmYhttps://voca.ro/14HoWChNdKhz[Intro]In darkest night, a light breaks throughEchoes of a dream, a call anew[Verse]"Ein Volk, ein Reich, ein Führer" I hearA promise of hope, a nation's fearRise and shine, a new dawn's rayA people's destiny, on a path astray[Chorus]Stay the course, through trials and strifeA vision of glory, a life of strife[Verse]A call to arms, a heart of goodA nation's fate, a leader's creedA people's rise, a world in need[Chorus]Stay the course, through trials and strifeA vision of glory, a life of strife[Outro]Through the storm, we'll find our wayTo a brighter dawn, a new day.
>>6151600Contribute something then shitstain.
not sure if AI
>>6151600>ai music>good you have to understand the aphantasia-afflicted third worlders who tend to use ai can't even be bothered to edit any of it to sound/look nice. i wouldn't even be surprised if half or more of the posts were first prompts with no actual human input.
>>6151831Scott Brown - Taking Drugs (2004)obviously coherent, competent and not ai even if edm is shit
>>6152030my space song is great.go and sign-in into youtube, do search for your pop-slop and stay there.
>>6152037The space song AI made and you posted blows monkey chunks lol.
The Cure: Just Like Heavenhttps://voca.ro/1cNt2KlzHQpUusing the original as audio source for timbre ref:https://voca.ro/12COn38PWmJIhttps://voca.ro/1miijoyXE5m3https://voca.ro/1elSJGnAZHHihttps://voca.ro/1mPBTgkEtzJ3
