Introduction to Spatial9

What Spatial9 does, who it is for, and how the platform, the engine and the integrations fit together.

Spatial9 is an AI platform that turns stereo audio, stems and existing immersive masters into immersive spatial audio, and delivers the result in open and industry formats such as Eclipsa Audio (IAMF), ADM BWF, binaural and Ambisonics. This page explains what the platform does, who it is designed for, and how the web app, the engine and the developer integrations fit together.

Read this page first if you are new to Spatial9. It gives you the mental model used by every other page in these docs: what you provide, what the system automates, what you still decide, and which entry point (the app, an AI assistant, an agent or the API) matches the way you work.

What Spatial9 does

Spatial9 places each sound in a mix at its own position around the listener, in front, behind, to the sides and above, instead of on the single left-to-right line that stereo allows. The platform recreates the cues the brain uses to locate sound, so the result can be heard on a room of speakers, a soundbar or ordinary headphones through binaural rendering.

The platform performs four kinds of work for you:

• Source separation. When you supply only a finished stereo master, Spatial9 estimates individual stems such as vocals, drums, bass and other instruments.
• Spatial arrangement. GravityLLM, the Spatial9 AI model, decides where every source sits, how far away it is, how high it is and how it moves over time.
• Processing and rendering. AI agents coordinate the processing and rendering steps that turn sources plus positions into immersive audio.
• Delivery. The system exports the deliverables your workflow needs, from binaural headphone files to ADM BWF masters and Eclipsa Audio for YouTube.

Who Spatial9 is for

Spatial9 is designed for anyone who has finished audio and wants an immersive version without building a dedicated immersive studio. Typical users fall into the following groups.

Design principles

The platform is built on a small set of principles that explain much of its behavior.

1. Accessibility for creators. Immersive production is available to independent musicians, video creators, labels and students, not only to large studios. A stereo track or a set of stems is enough to start.
2. AI does the heavy lifting. Separation, spatial arrangement, processing, rendering and multi-format export are labor-intensive tasks that the system performs quickly and consistently.
3. Humans direct taste and intent. The system proposes; you decide what the mix should feel like and when it is finished. Every move the engine makes can be heard, compared with stereo, edited or rejected.
4. Open formats and no lock-in. Spatial9 delivers to open and widely supported formats, including ADM, Eclipsa Audio (IAMF), binaural and Ambisonics, depending on the workflow and plan.
5. Accessibility for listeners. The music experience includes a hearing accessibility profile so playback can adapt to the person listening.

Spatial9 is a member of the Alliance for Open Media, the organization behind IAMF, and debuted music rendered in Eclipsa Audio with Google at CES 2025.

The two lanes of work

Every Spatial9 project divides work between two lanes. Understanding this split helps you know where to spend your attention.

The human lane

You act as the creative director. The human lane is a repeatable loop of five decisions:

1. Write the creative brief. Describe the genre, mood, destination and intent. A specific brief such as "intimate acoustic ballad for headphones, vocal close and centered, room ambience gently above" gives the system more to work with than a generic request.
2. Judge the source. Decide whether the stereo master is clean enough, whether stems are available, and whether problems such as heavy distortion or a very narrow mix will limit the result.
3. Listen critically. Review on headphones and, where possible, on speakers. Check balance, clarity, translation and whether the space serves the song.
4. Decide what to revise. Write actionable notes, for example "the vocal feels too far away" or "the hi-hats move too much".
5. Approve delivery. Check formats, file names and destination specifications before anything is published.

The automation lane

Spatial9 handles source separation when needed, generates the spatial arrangement with GravityLLM, coordinates processing and rendering with AI agents, and produces exports. For catalogs, jobs can be submitted in batches through APIs, with webhooks reporting back when a job finishes.

The production pipeline

A track moves through seven stages from upload to publication. These stages are the same whether you work in the app, through an AI assistant or through the API.

A separate training lane teaches GravityLLM how to make spatial decisions. Studio sessions contribute stems, spatial positioning data and genre metadata; a training database stores 3D coordinates, instrument types, genre classifications and movement patterns; and the trained model is versioned, tested and deployed to inference engines. Because spatial decisions are learned from real examples, the model places an electronic synth differently from a folk guitar. See Spatial inference and GravityLLM for details.

What the engine provides

The music engine combines several systems that respond to the structure and content of a song. The following table summarizes the capabilities documented for the engine.

The lead vocal, kick and bass stay anchored by default, so the song keeps its center while the space around it moves. See Spatializer for Music for the full description of these controls.

Products

Spatial9 is organized into the following products. Each has its own page in the Products section of these docs.

Integrations for developers and agents

The same engine is reachable outside the web app through several interfaces. All of them submit the same kind of render work and are subject to the same account limits.

The documented limits are 3 concurrent renders, 20 renders per day and 120 A2A tasks per hour. In the A2A interface, messageId acts as the idempotency key, and push notifications carry the X-A2A-Notification-Token header. See Limits, quotas and errors before you build automated pipelines.

Output formats

Depending on the workflow and plan, Spatial9 delivers the following families of output.

• Eclipsa Audio (IAMF). The open, royalty-free format based on the Alliance for Open Media IAMF specification. YouTube has accepted Eclipsa Audio uploads since January 2025.
• ADM BWF and ADM XML. Object-based masters in the Audio Definition Model, used for professional delivery workflows including Dolby Atmos.
• Binaural. A headphone rendering as WAV or MP4 that plays on any pair of headphones.
• Ambisonics. Scene-based audio (higher-order Ambisonics), used for VR and 360 video.
• Channel layouts. Stereo, 5.1, 7.1, 7.1.4 and larger layouts on plans that include them.

For input, the platform accepts WAV, MP3, FLAC and AIFF. For best results, use 24-bit WAV at 48 kHz or higher; files up to 96 kHz sample rate are accepted.

Plans at a glance

Music and video features are offered in separate plan families. Format availability depends on the plan.

Video plans start with Video YouTuber at $14 per month, and a Creator Bundle combines Music Pro and Video Creator. Current prices and yearly rates are listed on the pricing page.

How these docs are organized

Terms such as object, bed, scene, renderer and binaural are defined in Core concepts and in the glossary.

Next steps

• Quickstart: create, review and export your first immersive mix.
• Core concepts: learn the vocabulary used throughout these docs.
• Choose the right workflow: decide between the app, MCP, A2A, the REST API and the Acoustic Digital Twin.
• ADM to Eclipsa Audio converter: publish an existing Dolby Atmos master on YouTube.