Skip to main content
Audio / music / SFX

Audio — MusicGen, Stable Audio, SFX workflows

Music and sound-effect generation with the right stack documented and ready to install.

Launch standard
Pay by the hour · no contracts · terminate anytime from the dashboard
What this is

Music and SFX generation has moved past toys. MusicGen-Large, Stable Audio Open, AudioLDM 2, MAGNeT — they all have different sample-rate, conditioning, and CUDA assumptions. The instance boots on a bare CUDA 12.1 base — nothing preinstalled — but we publish the working matrix plus FFmpeg/NVENC install steps so you can render to wav / mp3 / video with baked-in audio without leaving the instance.

Our value-add

Not 'upload a prompt, get a link' — real tools, real control, real code you can version.

Environment & recommended stack

  • Base image: nvidia/cuda 12.1 devel (Ubuntu 22.04)
    clean boot — install the stack below
  • MusicGen-Large (text-to-music)
    recommended — 3.3B params, up to 30s clips
  • Stable Audio Open
    recommended text-to-audio, short clips, permissive license
  • AudioLDM 2
    recommended text-to-audio / sound effects
  • FFmpeg + NVENC/NVDEC
    apt install — render audio to video or re-encode
  • librosa + soundfile + pedalboard
    recommended post-processing / DSP

You stay in control

  • Bring your own reference audio for conditioning
  • Full root to install Suno-API style wrappers or custom forks
  • Prepaid credit cap + auto-destroy deadline for batch render jobs
Example workflow

Generate a 15-second piece with MusicGen

# pip install audiocraft
from audiocraft.models import MusicGen
import torchaudio
m = MusicGen.get_pretrained("facebook/musicgen-large")
m.set_generation_params(duration=15)
wav = m.generate(["slow ambient synth pad with a soft arpeggio"])
torchaudio.save("out.wav", wav[0].cpu(), m.sample_rate)
Run inside the instance after `ssh root@<instance-ip>` or via the web terminal.
Launch options

Pick a GPU for audio / music / sfx

Recommended start
Standard — MusicGen / Stable Audio
MusicGen-Large and Stable Audio fit comfortably in 48 GB
$2.04/ hour
Launch with this config
Scale option
Heavy — long-form / batch
For longer clips or when you're generating at scale
$4.41/ hour
Launch with this config

Hourly rates shown are current target rates — the authoritative price is surfaced on the /launch page at the moment you provision. Billing is per 5 minutes of runtime.