Skip to content
Audio

ORB planner

a compact AI model for generating sound specs from text

2 min readupdated

This 5.10M-parameter decoder-only transformer turns a single-line prompt (such as a description of weather or sound effects) into a structured JSON object (Spec). Rather than synthesizing raw audio directly, it hands the output off to a deterministic DSP renderer in the ORB repo, which compiles the format into a 48 kHz WAV file. Trained from scratch on 300k synthetic pairs without pretrained weights or external audio datasets, it is built for UI developers, game devs, and sound designers who need programmatic control over sound effect generation.

Key features

  • Compact size — only about 5.10M parameters with a 1024-token context window.
  • Strict grammar — grammar-masked decoding guarantees a valid JSON config every time without failures.
  • Wide sound palette — supports 47 sound roles (from clicks and whooshes to footsteps and lightning strikes), 13 textures (wind, rain, city noise, and more), and 10 moods.
  • Agent integration — the repository includes an MCP server so you can expose sound generation as agent tools.

What you can do

  • Generate sound effects — create custom UI pops, whooshes, and ambient textures from text prompts for apps and games.
  • Plug into pipelines — use the model as an audio planner and tweak parameters via a structured JSON object.
  • Connect AI agents — hook up the MCP server to let AI agents automate soundscape creation.
  • Procedural rendering — pair the generator with static DSP algorithms to get clean 48 kHz audio without neural audio synthesis artifacts.

Example

from huggingface_hub import hf_hub_download
from orb.api import load_planner, generate, save_wav

path = hf_hub_download("marcosmaceo/orb-planner", "planner.safetensors")
hf_hub_download("marcosmaceo/orb-planner", "config.json")
tok, planner = load_planner(path)
out = generate(
    "6.00-second bright cue, wind bed, with whoosh @1.00 0.90s -6dB L2 #2a",
    planner=(tok, planner),
)
save_wav(out["audio"], "whoosh.wav")
print(out["spec"])
Read moreDetails
• Accuracy on grammar test prompts: 100% (500/500 and 2000/2000 samples).
• Accuracy on free-form text prompts: 58.4% (custom data) and 70.0% (blind set).
• Decoding errors: 0.
• Training: 40k steps on a single H100 GPU (300k synthetic rows dataset).
• License: Apache-2.0.
Discuss Telegram 0 33

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 0.0 GB · ready files0.5 GB
  • FP8 / INT8recommendedweights 0.0 GB0.5 GB
  • 4-bitweights 0.0 GB0.5 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 2060 · BF16 — from RTX 2060

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # marcosmaceo/orb-planner
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download marcosmaceo/orb-planner --local-dir ~/models/orb-planner
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to orb-planner
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

New posts about versions of orb-planner (quants, finetunes, LoRAs) will show up under the bell.

No known relatives of this model yet. All derivatives on Hugging Face

About the model

marcosmaceo/orb-planner
Author
marcosmaceo
Category
Audio
Released
Posted

On NeyroLearn

Hugging Face

330

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments