Publishing Ambisonics: 360 Video for YouTube

YouTube can render first-order Ambisonics with a 360-degree or VR video. As the viewer changes direction, the platform rotates the sound field to match the view. This gives you a practical way to publish an ambisonic mix and share it with a link.

Supported audio formats

YouTube accepts two first-order Ambisonics formats. Both use AmbiX with ACN channel order, SN3D normalization, a 48 kHz sample rate, and one audio track in the uploaded video file.

4 channelsAmbiX FOAUse the channel order W, Y, Z, X. All four channels rotate with the 360-degree view.

6 channelsFOA with head-locked stereoUse the channel order W, Y, Z, X, L, R. The final stereo pair stays fixed to the listener.

The six-channel format does not contain a separate stereo audio track. It is one multichannel track with the Ambisonics channels first and the head-locked left and right channels last.

World-locked and head-locked sound

World-locked sound turns with the scene. If a bird is behind the viewer, it stays in that part of the scene when the viewer looks elsewhere. Most diegetic sound belongs in the Ambisonics channels.

Head-locked sound does not move when the viewer turns. Narration and non-diegetic music often work better this way. To use head-locked audio, deliver the six-channel format and route the stereo signal to channels 5 and 6. If you deliver four channels, every sound in the mix rotates with the scene.

Design and preview the full scene in Reaper

A normal video preview only shows the direction the viewer currently faces. The rest of the scene still needs sound. Treat the 360-degree video as one sound world with a shared front, back, left, right, top, and bottom.

Look around the video

Add Reaper’s Video Processor to the video track and choose the preset Panoramic: Equirectangular/spherical 360 panner. Its yaw, pitch, roll, and fov controls change the direction shown in the Video window.

Use yaw to check the four main directions:

Yaw	View
0 degrees	Front
90 degrees	One side
180 degrees	Back
-90 degrees	Other side

You can also bypass the spherical panner and work from the full equirectangular image. In this flattened view, the center is usually the front, the left and right edges meet behind the viewer, and the top and bottom show the vertical extremes. Confirm the orientation of the clip before placing sounds. Some cameras put the seam or default front in a different location.

Monitor the listener’s turn with ATK

The Video Processor changes the picture, but it does not send its viewing angle to ATK. Match the movement manually while you work. On a four-channel Ambisonics monitoring bus, use this chain:

ATK-encoded sound tracks
-> ATK FOA Transform RotateTiltTumble
-> ATK FOA Decode Binaural
-> headphones

RotateTiltTumble represents the listener turning inside the sound field. Turn the sound field in the direction opposite the video view. For example, when the video looks 90 degrees to the right, rotate the sound field 90 degrees to the left. ATK defines positive azimuth counterclockwise, so check a known sound at the front before relying on the signs.

Place each source in world coordinates with an ATK encoder such as Planewave. Do not rewrite its position for every possible view. A bird placed behind the starting view remains there. The monitor rotation only lets you turn toward it and hear the result from another direction.

A useful check takes less than a minute:

Play the section facing front at 0 degrees.
Check both side views at 90 and -90 degrees.
Check the rear view at 180 degrees.
Look up and down when the scene contains elevated sources.
Return the video view and RotateTiltTumble to 0 degrees.

The binaural decoder and this temporary listener rotation are for headphone monitoring. Do not bake them into the four-channel delivery. Before rendering, reset or bypass RotateTiltTumble, bypass the binaural decoder, and use the FuMa-to-AmbiX converter described below.

Convert ATK output to AmbiX

ATK processes first-order Ambisonics in FuMa B-format, while YouTube expects AmbiX. The formats use different channel orders and normalization. Uploading FuMa without conversion will distort the level and direction of the sound field.

Use the course FuMa-to-AmbiX JSFX after the ATK processors on the four-channel Ambisonics path. The ATK setup lesson shows how to install it. The converter changes FuMa W, X, Y, Z to AmbiX W, Y, Z, X with SN3D normalization; it has no format setting to select. For a four-channel delivery, render those four outputs. For a six-channel delivery, add the head-locked stereo mix to channels 5 and 6 after the converter, then render W, Y, Z, X, L, R.

Prepare and upload the video

Finish the ATK mix while monitoring through the room decoder or a binaural decoder.
Add the course FuMa-to-AmbiX converter after the ATK processors on the four-channel Ambisonics path.
Render one 48 kHz multichannel WAV. Use four channels for AmbiX alone or six channels when the project includes head-locked stereo.
Replace the video’s temporary audio with the multichannel file. For this lab, use PCM audio in a MOV container. YouTube also accepts four-channel AAC in MP4 or MOV and four- or six-channel Opus in MP4, subject to its bitrate and channel-mapping requirements.
Run Google’s Spatial Media Metadata Injector on the finished video. Select My video is spherical (360) and My video has spatial audio. The tool identifies head-locked stereo from the six-channel audio layout.
Upload the result as unlisted and wait for YouTube to finish processing it.

Check the uploaded version with headphones, in Chrome on the desktop or in the YouTube mobile app. Other browsers play a stereo downmix and will make correct metadata look broken. Drag the view to the right. A world-locked source should appear to move left relative to your head, while a head-locked source should stay in place. If the picture moves but the sound field does not, check the metadata. If directions are wrong, check the AmbiX conversion and channel order.

Listening list

You will not be able to record ambisonics of the scene in the video. That is the normal condition for this work. The pieces below built their sound fields entirely in post, from library material, foley, and synthesis, encoded to FOA and placed against the picture.

Play these in Chrome on the desktop or in the YouTube mobile app. Other browsers fall back to a stereo downmix without telling you.

Designed fields

Pearl . Google Spotlight Stories, 2016. An Emmy winner and the first VR film nominated for an Oscar. Nearly all of it happens inside one car, so the world-locked field is easy to follow as the view turns.
Sonaria . Directed by Scot Stafford, Google’s creative director for audio. Built with UC San Diego’s Sonic Arts department to get sources above and below the listener.
Behind the scenes: Sonaria . The team walks through the choice between head-locked stereo and the ambisonic channels. That is the same decision you make in the lab below.
Behind the scenes: Pearl .

Test material

Less interesting to listen to, more useful for calibrating what you are listening for.

Ambisonic audio test . One source, one rotation. Use it to confirm your playback chain works before you check your own upload.
Ambisonic audio test with 360 visualization . Draws the field on screen while it plays.
YouTube spatial audio test demo . Tones at twelve azimuth angles.
Wizard Battle 360 audio test . A fully invented field in third-order ambisonics.

Lab: publish one minute of ambisonics

Use a 360-degree video provided in class and material from your ambisonics project.

Place at least three sources against the picture. Include one stationary source, one moving source, and an ambience bed.
Add the course FuMa-to-AmbiX converter and render a four-channel 48 kHz WAV. If the project uses head-locked narration or music, render six channels instead.
Mux the audio with the video, inject the required metadata, and upload the result as unlisted.
Check the published video with headphones. Confirm that the world-locked field follows the scene and that any head-locked audio stays with the listener.
Exchange links with another student and check each other’s uploads before the end of class.