# 360 Degree Video Audio: Spatial Audio for YouTube 360 and VR > How to deliver spatial audio for YouTube 360 and VR video: first-order AmbiX (ACN/SN3D), head-locked stereo, SA3D metadata injection, how Eclipsa Audio relates, and how to check a file. Web page: https://spatial9.ai/blog/360-degree-video-spatial-audio-youtube Author: Spatial9 Team Published: 2026-11-03. Updated: 2026-11-03. ## Key takeaways - YouTube 360 and VR spatial audio uses first-order ambisonics in AmbiX format: ACN channel order (W, Y, Z, X), SN3D normalisation, 48 kHz. - Two layouts are supported: 4-channel FOA, or 6-channel FOA plus head-locked stereo for narration and music that should not rotate. - Only one audio track is allowed, and spatial metadata (an SA3D box) must be added, usually with Google's Spatial Media Metadata Injector. - Eclipsa Audio (IAMF) can carry ambisonic sound fields, but YouTube's documented 360 path is still AmbiX with SA3D metadata; use Eclipsa for flat immersive videos. - Check channel count, sample rate and metadata before upload; the free /try-your-files checker detects 360 metadata and ambisonic audio and offers a stereo vs binaural A/B. Short answer: for sound that turns with the viewer in a YouTube 360 or VR video, upload one audio track of first-order ambisonics in AmbiX format (ACN channel order, SN3D normalisation) at 48 kHz, either as 4 channels or as 6 channels with a head-locked stereo pair. Then add spatial metadata to the file, usually with Google's Spatial Media Metadata Injector. Without that metadata, YouTube treats the four channels as ordinary audio and the sound will not follow the picture. 360 video is the one place on YouTube where the viewer, not the editor, decides where the camera points. Sound has to keep up. If a waterfall is behind you when you turn around, you expect to hear it in front of you. This guide explains what YouTube asks for, why, and how to check a file before you upload it. ## How 360 audio differs from a normal soundtrack In a flat video, the frame never moves relative to the viewer, so a stereo or surround mix can be fixed to the screen. In a 360 video, the viewer drags the picture or turns their head with a headset or phone. The player then has to rotate the sound field by the same amount, in real time, on the viewer's device. That is why 360 audio is delivered as a sound field rather than as speaker channels. Ambisonics describes the sound arriving from every direction around one point, so the player can rotate it and then decode it to headphones. If you are new to the format, our ambisonics explainer covers the theory in plain English. ## What YouTube asks for YouTube Help lists the minimum requirements for spatial audio in 360 and VR videos. In summary: - Spatial metadata must be added to the file, either with the metadata tool or with post-production software that meets the YouTube specification. - Only one audio track may be used. A file with a spatial track and a separate stereo or mono track is not supported. - The audio must be ambisonics in AmbiX format: ACN channel ordering and SN3D normalisation. - The sample rate is 48 kHz. YouTube supports two layouts: | Layout | Channels in the file | Containers and codecs listed by YouTube Help | | First-order ambisonics (FOA) | 4: W, Y, Z, X | PCM in MOV; AAC in MP4 or MOV (at least 256 kbps); Opus in MP4 (channel mapping family 2, at least 512 kbps) | | FOA with head-locked stereo | 6: W, Y, Z, X, L, R | PCM in MOV; Opus in MP4 (channel mapping family 2, at least 768 kbps) | Note that the help page does not list AAC for the 6-channel layout. If you want head-locked stereo, PCM in MOV is the simplest safe choice. Requirements can change, so check the current help page before a delivery. ## AmbiX, ACN and SN3D in practice First-order ambisonics uses four components. W is the omnidirectional part. X, Y and Z are figure-of-eight patterns pointing front-back, left-right and up-down. - ACN (Ambisonic Channel Number) fixes the order of those components in the file: W, Y, Z, X. Google's Spatial Audio RFC gives the rule as n = l(l + 1) + m for degree l and order m. - SN3D (Schmidt semi-normalisation) fixes their relative levels. The older FuMa convention stores the same first-order components in a different order (W, X, Y, Z) with a different gain on W. A FuMa file uploaded as if it were AmbiX will place sounds in the wrong directions. Most ambisonic plugins and microphone converters offer an AmbiX output; choose it explicitly rather than trusting the default. ## Head-locked stereo: what it is for Head-locked stereo is a normal left and right pair that does not rotate when the viewer turns. YouTube Help describes it as audio "which doesn't change when a viewer moves their head". Use it for sound that belongs to the viewer rather than to the scene: - Narration and voice-over. - Score and non-diegetic music. - Interface sounds or anything that should feel "inside your head". Everything that exists in the filmed world, such as a speaker on camera, traffic or birdsong, belongs in the ambisonic part so that it stays put in the world. Our guide to moving camera, moving sound goes deeper into world-locked and head-locked decisions. ## Metadata: the step most uploads miss YouTube does not guess that a 4-channel file is ambisonic. Two pieces of metadata make a 360 video work: - Spherical video metadata tells the player the picture is 360 or 180 degrees. Google's Spherical Video V2 RFC defines it in an sv3d box. - Spatial audio metadata tells the player the audio is ambisonic. Google's Spatial Audio RFC defines an SA3D box inside the audio sample description, with fields for ambisonic order, channel ordering, normalisation and a channel map. A flag marks whether head-locked stereo is present. Google publishes the Spatial Media Metadata Injector in its spatial-media repository on GitHub. It can examine a file and inject metadata. The command-line form looks like this: ```text # Show any spatial metadata already in a file python spatialmedia my_video.mov # Inject 360 video and spatial audio metadata into a new file python spatialmedia -i --spatial-audio my_video.mov my_video_injected.mov ``` The README notes that the spatial audio option expects a first-order ambisonic track with ACN ordering and SN3D normalisation. Check the repository for the current Python requirement and for prebuilt desktop releases before you install it. Always write to a new output file; the tool will not overwrite its input. ## A typical workflow YouTube Help outlines a production path that most 360 projects follow: - Export the 360 picture as a proxy your audio software can play. - Open it in a DAW with a spatial audio workstation plugin set. - Place each scene sound in the sphere with the spatializer plugin, matching the picture. - Route head-locked material, generally stereo, to its own outputs. - Export the audio as WAV, then mux it with the video using the YouTube spatial audio export setting of your encoder. - Inject or confirm metadata, then upload. Before uploading, you can monitor the binaural result on headphones. YouTube Help points to the Resonance Audio Monitor VST plugin for this. Turn the view while listening: sources should stay attached to their picture, and head-locked material should stay still. ## How Eclipsa Audio relates to 360 audio Eclipsa Audio is the YouTube-facing name for IAMF, the Immersive Audio Model and Formats specification from the Alliance for Open Media. According to Google's open source blog, YouTube has accepted files with Eclipsa Audio (IAMF v1.0) since January 2025, with playback on a growing range of compatible devices. IAMF can carry channel-based audio elements such as 5.1.2 or 7.1.4 and scene-based audio elements, which are full ambisonic sound fields. So the format itself can hold the kind of sound field 360 video uses. That is not the same as saying YouTube uses IAMF for head-tracked 360 playback. The documented YouTube path for 360 and VR spatial audio is still first-order AmbiX with SA3D metadata. We have not found YouTube documentation that describes IAMF as a 360 delivery route, so we do not recommend relying on it for 360 projects. In practice: - 360 or VR video: deliver AmbiX FOA, optionally with head-locked stereo, plus metadata. - Normal flat video with an immersive mix: deliver Eclipsa Audio. See Eclipsa Audio for YouTube creators and the free conversion guide. ## How to check a file before upload Three quick checks catch most problems. ### 1. Count the channels Any media probe will show the channel count and sample rate. For example: ```text ffprobe -v error -select_streams a -show_entries stream=index,codec_name,channels,sample_rate -of compact my_video.mov ``` You want exactly one audio stream, with 4 or 6 channels at 48000 Hz. ### 2. Confirm the metadata Run the injector in examine mode, as shown above. It reports spherical video and spatial audio metadata if they are present. ### 3. Inspect and listen in the browser Our free file checker at /try-your-files reads the file in your browser, without uploading it. It detects 360 video metadata and ambisonic audio, explains what it found and offers a stereo versus binaural A/B so you can hear whether the sound field decodes as expected. It reads more than 50 formats and can also run a loudness, true-peak and clipping check based on ITU-R BS.1770. That figure is a guide for QC, not a certified meter. The broader checklist in is this file really immersive? covers the other signals, such as speaker masks and ADM metadata. ## Common mistakes - No SA3D metadata. The video turns, the sound does not. - FuMa instead of AmbiX. Sources appear in the wrong place, often mirrored or swapped. - Two audio tracks. A spatial track plus a stereo fallback in the same file is not supported. - Wrong sample rate. YouTube Help specifies 48 kHz. - Narration in the ambisonic mix. The voice moves around the viewer when it should stay put. - Checking on one device only. Test playback on a phone or headset as well as a desktop browser, and recheck after processing has finished. ## Scope of this guide This article summarises public documentation from YouTube and Google's spatial-media project as of November 2026, and how our own file checker handles 360 files. It is not a YouTube support document, and it does not cover other platforms' 360 specifications or higher-order ambisonics delivery. We have not measured playback behaviour across devices for this article. ## Frequently asked questions ### Does YouTube support higher-order ambisonics for 360 video? The YouTube Help page lists first-order ambisonics only, as 4 channels or 6 with head-locked stereo. Google's Spatial Audio RFC describes higher orders in general, but the help page is the delivery reference to follow. ### Can I upload a binaural stereo file for a 360 video? You can, but it will play as fixed stereo. Binaural audio is already rendered for one head direction, so it cannot rotate when the viewer turns. Our binaural vs spatial audio article explains the difference. ### Do viewers need headphones? Head-tracked spatial audio is designed for headphones, which is where binaural rendering works best. On speakers the experience varies by device; see headphone vs speaker spatial audio. ### Can Spatial9 make 360 audio for me? Spatializer for Video, which matches immersive sound to picture, is in early access. If your file is a 360 video, the checker at /try-your-files will tell you what it contains and point you to early access. ## Sources - Use spatial audio in 360-degree and VR videos, YouTube Help. - Upload 180 or 360-degree videos, YouTube Help. - Spatial Audio RFC, Google spatial-media project on GitHub. - Spherical Video V2 RFC, Google spatial-media project on GitHub. - Spatial Media Metadata Injector, Google spatial-media project on GitHub. - Resonance Audio Monitor VST: getting started, Resonance Audio. - Immersive Audio Model and Formats (IAMF) specification, Alliance for Open Media. - Introducing the open source DAW plugin for Eclipsa Audio, Google Open Source Blog. - ITU-R BS.1770: Algorithms to measure audio programme loudness and true-peak audio level, International Telecommunication Union.