Release Notes

2022.1 Highlights

An all new Full Face Neural Network with Audio2Emotions and a Full Face Character Setup including eyes and tongue.

Audio2Face is a combination of AI based technologies that generates full facial motion and lip sync that is derived entirely from an audio source. With new features such as Audio2Emotion, which analyses the audio and automates the animation of emotions into the characters performance. The new Full-Face Neural Network animates all the features of the Face including eyes and tongue. The new Character Setup feature also allows the user to setup Eyes, Teeth and Tongue in the Character Retargeting process allowing for a full Face character Transfer.


2022.1 New Features and updates

Full Face with Emotions

The new Full-face Neural Network has been trained with a range of emotions like joy, amazement, anger, and sadness.


Audio2Emotion is a Neural network that detects the emotion of a character from an audio performance and automatically keyframes the performance.

Full Face Character Setup

The Character Setup UI aids the user in setting up their character meshes and connecting them to a Full-Face Pipeline.

Key Framing

A simple Key framing UI is provided in the Emotion Panel and the Post-Processing Panel.

Mesh Tools

New features such as “Mesh Separate”, “Mesh Extract by Subsets” and Freeze Transformations are available to help setup your mesh for A2F.
  • Update kit sdk to 103.1

  • Stamp mesh UV map fix. (used in charTransfer and toolbox)

2021.3.3 Highlights

Preview Feature - Batch Audio Processing

Audio2Face 2021.3.3 is now available on Omniverse Launcher. Included in this update is a preview release of batch audio processing and some additional bug fixes. Batch processing will allow the user to export blend weight animation files for each audio file located in the Audio source directory as either .Json or .USD. .. image:: /content/images/A2F2021_3_3_Feature.jpg


  • BlendShape Generation

  • Streaming Audio Player

  • Update kit sdk to 102.1.2

2021.3.2 Highlights

BlendShape Generation and Streaming Audio Player.

Audio2Face 2021.3.2 brings two new features to the application. Blendshape Generation and a Streaming Audio Player. The Blendshape Generation tool allows the user to generate a set of Blendshapes from a Neutral Head mesh. Connect your character to Audio2Face using the Character Transfer process and the BlendShape Generation tool will do the rest. The shapes can be used as generated - or bring the meshes to your favorite DCC where you can edit the shapes to suit your need. In addition we have included a Streaming Audio Player. The Streaming Audio player allows the streaming of audio data from external sources such as Text to Speech applications via the gRPC protocol. Video tutorials can be found at NV On Demand for both new features.



  • BlendShape Generation

  • Streaming Audio Player

  • Update kit sdk to 102.1.2

  • fixed initial audio path for ov-url

  • Set a2f to support ov server for file IO with server type omniverse

  • blendshape generation layout for Z-up scene

  • Wizard: different layouts for different scenes

  • TTS window within Streaming demo scene

  • Support of abs net paths in core config

2021.3.1 Highlights

Export to Metahuman and Blendshape Solver symmetry option.

Audio2Face 2021.3.1 release brings important updates to the Blendshape conversion process by including a “pose Symmetry” option and the much anticipated support for Epic Games, Unreal Engine 4 - Metahuman. Support is provided via the Omniverse Unreal Engine Connector version 103.1 which can be found in the OV launchers “exchange” tab and under the “connectors” list. To see the entire end to end process - From Audio2Face - to setup within UE4 and Metahuman - please watch the video tutorial found at NV On Demand This is initial version supports MetaHuman via the Blendshape conversion process and export to USDskelanim format. The UE4 connector will allow you to import your audio2face generated animation to setup on a UE4 Metahuman character.



  • Blendshape Solve: Symmetry Option - Allows the user to determine how much symmetry the blendshape solve should have.

  • Save / Load Blendshape conversion preset from OV server

  • New “delete setup” button for deleting the Blendshape solve setup

  • USD export adds UV support

  • Export all indices

  • Splash screen image fix

  • Maya Connector blendshape export fix

  • Audio2Face to Metahuman

  • Metahuman animation curve export from Blendshape conversion

  • Requires Omniverse Unreal Engine Connector version 103.1 to support the import of A2F data



  • A2F Data Conversion Tab: With the addition of Blendshape support - We have consolidated all of the export processes to one tab dedicated to data output.

  • Blendshape Support: Provides the functionality to connect a blendshape driven asset - to our A2F asset

  • A2F Attributes - Save/Load: The ability to save the current state of A2F in a USD file.

  • Kit SDK 102.

  • Geom Subset Deformation: Now supported.

  • Wireframe toggle Menu in Rendering > Visualizers >> Toggle Wireframe

  • No longer need omni.hydra.ui (setting is on Edit>Preferences>UsdSkel section), Blendshape is default enabled.

  • Prox Deformer. Animation > Deformers > Show PROX UI. This is similar to wrap, but has option for rigid partition bind for hair cards.

  • Latest Toolbox. Added the BUILT-IN > SkelAnimation > Create BlendShapes ROM

  • Mesh menu. Mesh separate tools by GeomSubset or partition.

  • New A2F icon

  • Updates to Blendshape export options

  • Prox deformer Bug fix.

  • Added precacher.


  • New Default Layout : to provide easier access to the properties panel and a more efficient workflow.

  • Character Transfer Presets: Allows the user to save presets for the setup of the character transfer process.

  • Reallusion Presets: Included are Presets for Reallusion Characters.

  • Presets: save and load from nucleus server path

  • Multi mesh export: ability to export multiple meshes in a single export operation.

  • Auto mesh selection when importing template and when selecting A2F mesh.

  • TRT encoding fix

  • Warning if mesh has disconnected points

  • Filter pickable points to not contain disconnected points

2021.2.5 Hotfix

  • Added Tooltip

  • Auto transform for DeltaBlend is now fixed

  • Full Undo Wrap UI is now supported

  • Delete multiple Transfer Setups

  • Terminal Node Performance is fixed and can now be enabled/Disabled.

  • Fixed mesh fitting window and not showing up when there is no Default Prim.

  • Correspondence creation undo.

  • Recent Files, Bookmarks, and other history related interface elements now work as expected.

  • Mesh Fitting is now fixed for meshes containing points without connections.

  • Viewport draw nodetype updated

  • omni.audio2face.node_updater extension which update nodetype automatically. Updated file is not backward compatible!

2021.2.0 Open Beta

  • Char Transfer Delete

  • Wrap UI fix

  • Default export dirs

  • New Kit SDK 101.288

  • Audio normalization

  • Recorder and live mode

  • A2F panel update

  • New xform node

  • A2F+player in one click

  • UI updates, minor bug fixes

  • Split app to wizard+tool

  • Wrap UI

  • Major refactor

  • Add char transfer

Known Limitations

  • Delete Correspondence button only works on single specified setup.

  • It’s only possible to run 1 instance of Audio2Face.

  • Tongue correspondence is not adjustable at this time.

  • User will need to save and reopen the scene after un-hiding correspondence meshes if they are previously hidden at scene open.

  • Audio cannot currently be accessed from omniverse Servers. Hotfix coming soon.

In some situations A2F may fail to load due to conflicts. Try the following if A2F fails to load correctly.

  • If a user has torch installed by another Kit app (not just pure torch, but from a Kit app), then it will conflict with A2F (due to cuBLAS)

  • In A2F, omni.warp loads torch even if it is not used.

  • To Remove Torch Look for this pip3-env folder and remove it.

  • C:Progra C:UsersUSERNAMEAppDataLocalovdataKitAudio2Face2022.1pip3-envs

  • C:UsersUSERNAMEAppDataLocalovdataKitAudio2Face2022.1pip3-envs

  • We are following up with a hotfix shortly.