Abstract: Non-player characters (NPC) in a computer game may be controlled using context in the form of player statements from the game engine to an LLM to drive NPC dialog and control NPC actions. Imperative statements are executed by game engine, and the result of this execution plus the original player request are converted to plain text description and injected back into the NPC dialog to ensure the NPC remains within defined bounds.

Abstract: Instead of feeding data pairs “input”-> “output” into a LLM to train it, an artificial “output”-> “input” dataset is generated to fine tune the LLM on it. As an example, it may be difficult to parse object/action words from a random sentence. But generating a random sentence given an object, action pair can be solved easily, and these reversed pairs are used to fine tune the LLM.

Abstract: Techniques are disclosed herein for providing audio enhancement and optimization of an immersive audio experience. Examples may include generating an audio feature set for a transduced audio stream captured in an environment, inputting the audio feature set to a neural network model configured to generate an audio isolation mask associated with the transduced audio stream, and generating isolated audio for the transduced audio stream based at least in part on the audio isolation mask.

Abstract: Techniques for providing an artificial intelligence denoiser related to audio processing are discussed herein. Some embodiments may include providing an audio signal sample associated with at least one microphone to a time-frequency domain transformation pipeline for a transformation period. Some embodiments may include providing the audio signal sample to a deep neural network (DNN) processing loop that is configured to determine a denoiser mask associated with a noise prediction for the audio signal sample. In a circumstance where the denoiser mask is determined prior to expiration of the transformation period, some embodiments may include applying the denoiser mask associated with the noise prediction to a frequency domain version of the audio signal sample associated with the time-frequency domain transformation pipeline to generate a denoised audio signal sample associated with the at least one microphone.