Production Software
The Leading Automatic Speech to Facial Animation Solution
SGX is an award winning, production tested, automatic speech-to-facial animation system. Over 20 years of R&D in speech technology, linguistics, machine learning and procedural facial dynamics are at the core of SGX. Through the innovative application of acoustic analysis and muscle dynamics, SGX produces accurate, consistent, natural lip sync and remarkable full-face emotional expression from just audio and text.
KEY FEATURES
- Mass Scale Ready – Generate tens of thousands of animations in minutes
- Full face Animation – Generate life like lip sync animation with audio driven full facial expressions
- Anatomically Based – Speech is analyzed and translated into virtual muscle activations within our vocal tract simulation
- Completely Automatic – No manual alignment of facial poses or visemes to audio clips
- No runtime – Playback the keyframe data like you would any other animation
- Easy Pipeline Integration – Maya plugins make it easy to add SGX to nearly any animation pipeline
- Scriptable – Completely automate your facial animation production
- Multi-language support – Localize your animation automatically
- Works with any 3D character – Compatible with hybrid, blendshape, and joint based character animation rigs
“The team at Respawn used Speech Graphics to generate thousands of high-quality facial animations for gameplay lines in Star Wars Jedi: Fallen Order . By analyzing our audio and written transcripts of our dialogue, the Speech Graphics software produced animations at a high fidelity to both the sound and the actual emotion of the voice acting. We were immediately impressed by how the whole face of the in-game character would reflect the tone of the line — even before adding the emotional markup that Speech Graphics offers. Thank you!”
David Midgley
Senior Technical Designer
SGX USE CASES
Enabling the next-generation of blockbuster aaa and indie titles.
- Automatically animate hundreds of thousands of lines of dialog
- Build worlds teeming with speaking NPCs
- Tell even bigger stories and polish cutscenes
- Localize animation to match localized audio
- Add emotion and better connect with characters in any genre
- Free animators to work on other critical tasks
TRAINING + EDUCATION
Improve learning and engagement with compelling characters.
- Animate deep dialog trees to simulate customer conversations
- Tutorials, and user guides
- Interactive installations, and kiosks
FILM + TELEVISION
Increase animator output.
- Accelerate speech animation for short form, episodic, and feature animation
- Automatically generate animation for markets around the world from localized audio
- Add emotional performances to Pre-Viz
“Speech Graphics provided efficient, quality results for our automated in-game speech. The Maya tools and plug-in allowed us to easily iterate on the results of the audio analysis onto our custom face rig until we met our visual target, without the need for animators to spend manual effort on individual face animations. I’m very happy with the results, and would recommend Speech Graphics as a solution for any audio based, automated face animation pipeline.”
David Coleman
Technical Art Director, XBox Game Studios – The Coalition
CREATE EMOTIONAL ANIMATION
Direct the emotions of your characters’ animations using Mood Markup.
When preparing your transcripts for processing, define the desired mood and intensity to realize the emotional nuance you envision. With the included SG Studio Maya plug-in you can create an unlimited library of custom moods that are exactly right for your characters and dialog.
- Mood library – Author and choose from an unlimited number of moods with an unlimited number of facial expressions for a given mood
- Bi-Tonal Moods – These special moods modulate their expression based on changes in the pitch and intensity of audio
- Markup – Set the mood for a performance across a batch of lines, for an individual line, or change moods multiple times within a single line of dialog
- Emotion Detection – Automatically modulate the mood of a character based on the Positivity, Negativity, and Neutrality of vocal performance
THE SGX WORKFLOW
Begin to animate a single line of dialog or a thousand lines in just a few clicks using our GUI or by launching a custom script.
“Speech Graphics’ SGX enabled the team at Eidos-Montréal to generate over 20,000 high quality lip-sync animations for Shadow Of The Tomb Raider with its wide range of conversations. The generated content integrated directly into our existing animation pipeline, which made it a nice addition to our tools and the team provided fast and expert support to our needs throughout our production.”
Pierre-Marc Simard
Animation Technical Director, Eidos-Montréal
MULTI-LANGUAGE SUPPORT
SGX can generate life like emotional facial animation in twelve languages. By localizing your audio AND your animation you can better connect with audiences in markets around the world.
SUPPORTED LANGUAGES
- English (UK, USA)
- Spanish (European, Latin America)
“The Speech Graphics tools easily integrated into our automated build system, and gave us deterministic results to allow for minimal incremental builds. We distributed execution of the tools in parallel to generate animations for over 100,000 speech lines for Gears 5 . I’m very happy with the performance and results from the Speech Graphics command line tools and Maya plugin.”
Principal Software Engineer, XBox Game Studios – The Coalition
LICENSING OPTIONS
SGX has a number of license options including; Indie, Production, and Studio. Please contact us to schedule a consultation. We look forward to talking soon and finding the right SGX license for you and your project.
You’re in good company
- Speech Graphics Knowledge Base
What's New in SGX 4.3?
The SGX 4.3 suite includes:
SGX Production Tools 4.3
SGX Studio Maya 4.3
SGX Studio Unreal Engine 4.3
See release notes for a full list of improvements and fixes. Here are the highlighted features:
Improved lip sync
Improved expression intensity, improved language modules, dual animation levels, muscle activation display.
https://vimeo.com/996105124/55d7385f39?share=copy
SGX now animates breath!
This is more than a simple breathing loop. SGX detects the sounds of breathing in the input audio, and the breath animation is synchronized to the audible inhales and exhales. Breath is also woven naturally into the pauses in speech. The intensity, speed and rate of breath are all determined by the vocal performance.
In order to make use of automated breath animation, the only requirement is to have the new Chest Breath muscle defined.
Below are two examples of automatic breath animation: panting breath, and breath mixed with efforts and speech.
https://vimeo.com/993570896/ab00deee64?share=copy
https://vimeo.com/993572361/4ebef3f7b6?share=copy
Effort with breath
If desired, breath behavior can be adjusted by two new Behavior Modifiers: Breath Magnitude and Breath Speed .
In addition to animating breath, SGX also outputs breath metadata , in the form of inhale and exhale segments. This metadata can be edited to change breath timing, and is included in metadata export .
As part of our mission to automate animation for all of the vocal sounds a human can make, SGX 4.3 adds a new auto mode : Laugh . This is triggered whenever laughter is detected in the audio. The behavior of the new auto mode is defined by mapping it to a behavior mode with various laugh-like expressions. Below are a couple of examples of automatically generated laughter animation using the Laugh auto mode.
https://vimeo.com/993577666/f6722b6763?share=copy
https://vimeo.com/993579216/f9ce087102?share=copy
Evil laughter
As of SGX 4.3, an algorithmic breakthrough in phonetic lip sync provides improvements in alignment of speech motion with the voice. The timing improvements are especially noticeable for softer-edged sounds like glides and liquids. These improvements apply to all languages.
You should notice generally improved expression intensities. That is, the intensity of the facial expressions should better match the intensity of the vocal performance.
The language modules have been optimized:
Each module now consists of just a single file. For example, English_US_2.2.k no longer has an auxiliary folder English_US_2.2.k_ . This streamlining makes the languages easier to download and install.
Languages load significantly faster, which cuts down on processing time.
Users can now download language modules directly from the Downloads site, rather than via delivery.
In SGX Director, the language dropdown will only display the user’s licensed languages.
Note that SGX 4.3 is not compatible with older language modules, so you must download the latest.
Character rigs can have two animation levels - high-level UIs such as control rigs, and low-level parameters such as bones, blendshapes or anim curves. We call these the control level and the baked level , respectively. In typical cases, the control level is better suited for manually setting up the muscle poses, but the baked level is used for animation in the game engine.
A Character Control File that animates the control level can be retargetted to the baked level. Previously, this retargetting process produced a second ‘baked’ Character Control File. This resulted in two Character Control Files for a single character, and any changes made to the original file required re-generating the second file. This situation often led to confusion between the two files and situations where they became out of sync.
In SGX 4.3, a single Character Control File can now contain both animation levels. All that is required to make use of this feature is to select a second set of baked targets, and call a function to bake the muscles. See Baked animation level on how to use this feature in SGX Studio.
During playback of an event in SGX Director, you can now view the state of the muscles in an instance of SGX Studio bound to the event. The muscle sliders are updated whenever playback pauses or the play position is scrubbed.
Please note, these errors can depend on your browser setup.
If this problem persists, please contact our support.