New #1 open-source AI music generator is INSANE!
Video Overview & Insights
Ace-Step 1.5 tutorial. How to install Ace Step 1.5. Best free AI music generator. #ai #aimusic #aitools
NOTE: A BETTER VERSION IS OUT! See my XL tutorial here: https://youtu.be/UAlLD5fS7-c
Thanks to our sponsor LumaLabs. Visit https://lumalabs.ai/aisearch10 or scan the QR code on screen to try The Dream Machine for free
NOTE: A BETTER VERSION IS OUT! See my XL tutorial here: https://youtu.be/UAlLD5fS7-c
Thanks to our sponsor LumaLabs. Visit https://lumalabs.ai/aisearch10 or scan the QR code on screen to try The Dream Machine for free
1) Online AI tools were so famous.
2) Offline AI will kill all those giant online AI.
Ace Step 1.5 github: https://github.com/ace-step/ACE-Step-1.5
Git https://git-scm.com/install/windows
imho i'm still gonna use suno until local models can handle (echoing) and [vocal instructions] till then i'll see you guys in the black
ComfyUI workflow: https://blog.comfy.org/p/ace-step-15-is-now-available-in-comfyui
ALSO SEE HeartMula tutorial: https://youtu.be/54YB-hjZDR4
tried it and it not that good for complex stlying
0:00 Ace Step 1.5 intro
1:35 Rock demo
Music for poors.
2:29 Indian fusion
3:18 Country folk
A few questions: 1) Why does my UI look different to on the video? I have no "task type" and "required input" section .. 2) Everytime I open the interface, it reverts to MP3 128 kbps when I want it at 320 kbps (or FLAC), how do I save this by default? 3) After 2 generations, I always run out of VRAM and have to restart the whole thing. How do I get it to flush the VRAM after each generation? (I have 24GB RTX 4090)
3:50 Kpop
5:33 Spanish reggaeton
If that's a metal song, then the metal scene is fucked.
6:56 Jazz lofi
7:38 Jpop EDM
Please the Lora training!
8:21 Chiptune instrumental
9:10 Vocal chops
really wanted to hear a rapping example. Ai is so bad at that because it usually defaults to "Pop" singing so it ends up sounding so goofy on a rap song.
9:37 Epic orchestra
10:18 Salsa
Well, if you are reading this: Don't waste your time. These are REALLY trash models.
The input: A aggressive metal song. The result: A eletronic house music.
11:00 Ace Step 1.5 specs
11:49 How to install Ace Step 1.5
I am so excited about plastic music!!!
13:02 Git
13:49 Installation continued
Boom boom tel aviv
15:45 How to run Ace Step 1.5
16:47 Ray 3.14
Anyone help please:
❌ Error initializing model: Error no file named pytorch_model.bin, model.safetensors, tf_model.h5, model.ckpt.index or flax_model.msgpack found in directory C:\Users\latoa\Desktop\ACE-Step-1.5\checkpoints\acestep-v15-turbo.
Traceback:
Traceback (most recent call last):
File "C:\Users\latoa\Desktop\ACE-Step-1.5\acestep\handler.py", line 404, in initialize_service
self.model = AutoModel.from_pretrained(
^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\latoa\Desktop\ACE-Step-1.5\.venv\Lib\site-packages\transformers\models\auto\auto_factory.py", line 597, in from_pretrained
return model_class.from_pretrained(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\latoa\Desktop\ACE-Step-1.5\.venv\Lib\site-packages\transformers\modeling_utils.py", line 277, in _wrapper
return func(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\latoa\Desktop\ACE-Step-1.5\.venv\Lib\site-packages\transformers\modeling_utils.py", line 4900, in from_pretrained
checkpoint_files, sharded_metadata = _get_resolved_checkpoint_files(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\latoa\Desktop\ACE-Step-1.5\.venv\Lib\site-packages\transformers\modeling_utils.py", line 989, in _get_resolved_checkpoint_files
raise OSError(
OSError: Error no file named pytorch_model.bin, model.safetensors, tf_model.h5, model.ckpt.index or flax_model.msgpack found in directory C:\Users\latoa\Desktop\ACE-Step-1.5\checkpoints\acestep-v15-turbo.
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "C:\Users\latoa\Desktop\ACE-Step-1.5\acestep\handler.py", line 415, in initialize_service
self.model = AutoModel.from_pretrained(
^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\latoa\Desktop\ACE-Step-1.5\.venv\Lib\site-packages\transformers\models\auto\auto_factory.py", line 597, in from_pretrained
return model_class.from_pretrained(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\latoa\Desktop\ACE-Step-1.5\.venv\Lib\site-packages\transformers\modeling_utils.py", line 277, in _wrapper
return func(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\latoa\Desktop\ACE-Step-1.5\.venv\Lib\site-packages\transformers\modeling_utils.py", line 4900, in from_pretrained
checkpoint_files, sharded_metadata = _get_resolved_checkpoint_files(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\latoa\Desktop\ACE-Step-1.5\.venv\Lib\site-packages\transformers\modeling_utils.py", line 989, in _get_resolved_checkpoint_files
raise OSError(
OSError: Error no file named pytorch_model.bin, model.safetensors, tf_model.h5, model.ckpt.index or flax_model.msgpack found in directory C:\Users\latoa\Desktop\ACE-Step-1.5\checkpoints\acestep-v15-turbo.
❌ Error initializing 5Hz LM: Error no file named pytorch_model.bin, model.safetensors, tf_model.h5, model.ckpt.index or flax_model.msgpack found in directory C:\Users\latoa\Desktop\ACE-Step-1.5\checkpoints\acestep-5Hz-lm-1.7B.
Traceback:
Traceback (most recent call last):
File "C:\Users\latoa\Desktop\ACE-Step-1.5\acestep\llm_inference.py", line 223, in _load_pytorch_model
self.llm = AutoModelForCausalLM.from_pretrained(model_path, trust_remote_code=True)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\latoa\Desktop\ACE-Step-1.5\.venv\Lib\site-packages\transformers\models\auto\auto_factory.py", line 604, in from_pretrained
return model_class.from_pretrained(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\latoa\Desktop\ACE-Step-1.5\.venv\Lib\site-packages\transformers\modeling_utils.py", line 277, in _wrapper
return func(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\latoa\Desktop\ACE-Step-1.5\.venv\Lib\site-packages\transformers\modeling_utils.py", line 4900, in from_pretrained
checkpoint_files, sharded_metadata = _get_resolved_checkpoint_files(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\latoa\Desktop\ACE-Step-1.5\.venv\Lib\site-packages\transformers\modeling_utils.py", line 989, in _get_resolved_checkpoint_files
raise OSError(
OSError: Error no file named pytorch_model.bin, model.safetensors, tf_model.h5, model.ckpt.index or flax_model.msgpack found in directory C:\Users\latoa\Desktop\ACE-Step-1.5\checkpoints\acestep-5Hz-lm-1.7B.
18:44 Downloading AceStep models
22:16 How to initialize
Sorry this is in mode of SUNO version 3,5, the new Version of SUNO V5.5 is another level!
24:33 Text to song
29:07 How to repaint with Ace Step 1.5
LORA TRAINING PL
31:19 How to cover songs
33:35 Lora training
I have 3090 but on sft and 4B thinking model i get
Error: Insufficient free VRAM: need ~3.6 GB, only 1.2 GB available. Reduce batch size (currently 1) or audio duration
how to fix this. I checked hardware monitor and no VRAM usage at that moment
34:15 ComfyUI workflow
Learn AI: https://ai-search.io/courses
upon generating got this error: MP3 export failed without fallback: ffmpeg executable not found. Install ffmpeg or add it to PATH to export MP3 files. i would appreciate the help
Newsletter: https://aisearch.substack.com/
Find AI tools & jobs: https://ai-search.io/
I have an Error. When I Generate Music. The error is ffmpeg executable not found. Install or add it to PATH to export MP3 files. Please help
Support: https://ko-fi.com/aisearch
Here's my equipment, in case you're wondering:
From an actual music producing point of view, is it useful?
I´ve seen ACE Studio and the tools to create layers of for example piano or strings for your own track, in correcxt bpm and key ofc.. Is that possible with ACE Step yet? Can I just create a single layer of instrument for my existing track?
Lenovo Thinkbook: https://amzn.to/4jWeKwH
Dell Precision 5690: https://www.dell.com/en-us/dt/ai-technologies/index.htm?utm_source=AISearchTools&utm_medium=youtube&utm_campaign=precisionai#tab0=0
I set it up and I am testing it currently. The problem that I find is taht the singing is not very clear and many times the lyrics are not correct. It uses none existing words. Also the quality of the instruments / music is not very high. Am I doing something wrong or was the "better than Suno" statement exegerated?
GPU: Nvidia RTX 5000 Ada https://nvda.ws/3zfqGqS
Mic: Shure SM7B https://amzn.to/3DErjt1
Suno is much better unfortunately.
Audio interface: Scarlett Solo https://amzn.to/3qELMeu
The interface is easy, the models are poor. And it absolutely ignores prompts. SUCH AS NO DRUMS OR DRUM KITS. Respect to the project but AI Search's big hype - that's about all this amounts to. I will try out some other open source model based projects.
More User Perspectives
finally some good ai music stuff
@dineshkhumanthem9308The genre switching is impressive, it’s like branching logic. Pneumatic Workflow already handles conditional paths for my team, and this made me see music prompts similarly.
@ashkitum54900this is wild, can't believe it's open source
@sayedalaa4620this is gonna change everything for my projects 🤯
@ShimonChouriyawhoa i need to try this out
@kid105hardware wasted, electricity wasted, time wasted, internet flooded with garbage!
no matter how many generations you do and how good result you get - it's not going to make any of you a musician if the only time you seen musical instrument is when ai model generated an image of one for you....
i hope you people do realize, that you are giving away your ability not only to create something authentic, but also an ability to think to a machine.
yeah train some pink Floyd , joe satch with some others no vocals would like to see the outcome
@eddyhazethe maximum length 30s ?
@RealDailyCraftI am sorry but this is no better than Suno, and this program is just good for 2010's onwards mainstream "plastified" music ultra post produced. IF you want to create something that needs realistic organic sound like a soundtrack, this engine simply sucks donkey balls.
@perrymanso6841it's not free
@albertoagazzani4826Has any body found or does anybody know if ACE STEP 1.5 has a TEMPO LOCK feature? What is that? If you set the projects tempo ( BPM ) to say 120bpm, after it renders the WAV and you add this wav to your DAW ( Digital Audio Workstation eg Reaper ), set the DAWs tempo to the 120BPM .. There you notice the Ace Step 1.5 tempo / bpm is NOT the same across the whole track. Yes you can use MELODYNE to sort this but that software is not cheap. I noticed recently SUNO sorted this by offering two WAV expert version - one as is and the other as WAV (Tempo Locked ).
This would be useful for those of us who use Ace step 1.5 to generate new track ideas and might stem seperate the WAV out put and add the seperate instrument / vocal wavs to our DAW to redevelop this track or use it as a base for a complete new track played by ourselves in our DAWs
Fuck NVIDIA.
AMD is better for Linux, with far longer driver support due to proper open source drivers.
bro im dying here at 3am trying to get this stupid ace 1.5 thing to work. the suno replacement music ai thing.
i had to move to WSL and for the last 3 hours its been a "FFmpeg + TorchCodec mismatch." ive reinstalled ffmpeg and all these other things and resinstalled pytorch vision audio torchcodec im losing my god hecking mind someone HELP MEEE
Doesn't work.......delete
@DarrenAllencan create a bachata song ?
@vincenzocalendini8231Only 4 languges
@shahshohag8737Can you generate instrumentals only
@yung_winnerGood but audio quality and mixing, mastering bad , suno free version is little bit quality, any ai have like a suno paid version quality?
@emoneylanka7699I was struggling with clunky web tools until I tried an iOS app called SwapStyle AI. It generates AI music covers and lets you swap song genres instantly right on your phone. Highly recommend it if you want to do this on mobile!
@bugrahanefe6469how do I move the entire installation to SSD storage?
@thejoshuachineduIsnt it normal that audio ai is always low ram.
@PatchworxStudiosThink I need to try re-installing this or running it different because with a 1060 GTX 6gb trying to run remix on a 10 second clip, I'm getting crashes and errors left, right and centre... Has me wondering if I can run this at all...
@MikeSharp-y7lThank you friend
@AlphationUsing custom lyrics, and trying several models, and setting the right settings to adhere to my lyrics most strictly, I still can't get it to stick to my lyrics. It sometimes adds nonsense, and no single try has been OK. Unlike Suno which does exactly what I tell it to. Any tips for getting it to follow my lyrics 100%? It sure would be nice if this would work for me.
@galadrhim1whoa didn't expect it to sound this good
@Arthdwivedi-b4xthis is wild, i gotta try it out
@kamrankhan-vx5mqno way this is open source, crazy! 🎶
@WizKhalifaJkbeen waiting for something like this lol
@18zaid