Skip to content

akash-music-ai-generator

Directory actions

More options

Directory actions

More options

Latest commit

 

History

History

akash-music-ai-generator

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
Dockerfile
Dockerfile
 
 
README.md
README.md
 
 
deploy.yml
deploy.yml
 
 
studio-akash-final.html
studio-akash-final.html
 
 

README.md

Akash Music AI

AI-powered music generation on decentralized GPU compute. Deploy a fully functional music generation service to Akash Network with a modern web UI.

Features

  • ACE-Step 1.5 Music AI — State-of-the-art music generation from text descriptions
  • Decentralized Deployment — Run on Akash Network's distributed GPU compute
  • Custom Web UI — Modern, responsive interface for generating music
  • CPU Fallback — Graceful degradation if GPU VRAM is constrained
  • Multiple Output Formats — Generate MP3, WAV, or FLAC

Quick Start

Prerequisites

  • Docker installed locally
  • Akash CLI (akash) configured
  • Docker Hub account (or similar registry)

Deployment

  1. Build & Push Docker Image
docker build -t YOUR_DOCKER_USERNAME/acestep-api:latest -f Dockerfile . --no-cache
docker push YOUR_DOCKER_USERNAME/acestep-api:latest
  1. Update deploy.yml

Replace YOUR_DOCKER_USERNAME in deploy.yml with your Docker Hub username.

  1. Deploy to Akash
akash tx deployment create deploy.yml --from <your-key> --node $AKASH_NODE --chain-id $AKASH_CHAIN_ID
# Follow prompts to accept a provider bid
  1. Access the UI

Once deployed, navigate to your provider's endpoint:

http://<provider-ip>:<exposed-port>/

Customization

Custom UI

Edit studio-akash-final.html to customize colors, layout, or add features:

# Edit the HTML file
nano studio-akash-final.html

# Rebuild the Docker image
docker build -t YOUR_DOCKER_USERNAME/acestep-api:latest -f Dockerfile . --no-cache
docker push YOUR_DOCKER_USERNAME/acestep-api:latest

# Redeploy using updated `deploy.yml`

GPU Selection

Modify deploy.yml to target specific GPU types:

gpu:
  units: 1
  attributes:
    vendor:
      nvidia: true
    model:
      a100: true  # or rtx4090, l40s, etc.

Pricing

Adjust the bid in deploy.yml to attract higher-end providers:

placement:
  akash:
    pricing:
      acestep-api:
        denom: uakt
        amount: 50000  # Higher = better GPUs

Architecture

┌─────────────────────────────────────┐
│    Browser (Web UI)                 │
└──────────────────┬──────────────────┘
                   │
                   ▼
        ┌──────────────────────┐
        │   Nginx (Port 80)    │
        │   - Serves HTML      │
        │   - Proxies /api/*   │
        └──────────────────────┘
                   │
                   ▼
        ┌──────────────────────┐
        │   FastAPI Server     │
        │  (Port 8001, local)  │
        │  - /release_task     │
        │  - /query_result     │
        │  - /health           │
        └──────────────────────┘
                   │
                   ▼
        ┌──────────────────────┐
        │  ACE-Step 1.5 Models │
        │  - DiT (Music Gen)   │
        │  - LM (Enhancement)  │
        │  - VAE (Codec)       │
        └──────────────────────┘

Tips for Success

  • Start with shorter durations (30-60 seconds) to avoid VRAM issues
  • Use descriptive prompts — "upbeat electronic dance music with heavy bass" works better than "good song"
  • Monitor logs — Check provider logs for generation status and any VRAM warnings
  • Upgrade if needed — If generation times are slow, deploy to a provider with RTX 4090, A100, or H100

References

License

This project wraps ACE-Step 1.5 — refer to their repository for licensing details.

Support

Issues or questions? Check the logs in your Akash deployment console or open an issue on GitHub.

Built with ❤️ on Akash Network