PUBG Ally: KRAFTON co-playable AI built with NVIDIA ACE

KRAFTON built PUBG Ally, a co-playable AI teammate for PUBG: BATTLEGROUNDS, and the system is powered by NVIDIA ACE and its suite of efficient models and tooling. PUBG Ally uses automatic speech recognition, a 2B-parameter small language model, and text-to-speech to understand and respond to players.

What is PUBG Ally and who made it?

PUBG Ally is a co-playable AI character created by KRAFTON for PUBG: BATTLEGROUNDS, and it is powered by NVIDIA ACE. KRAFTON built the system specifically for PUBG: BATTLEGROUNDS, and NVIDIA ACE supplies the models and tooling that power the teammate’s perception and speech pipeline.

KRAFTON positioned PUBG Ally as an AI teammate integrated into the game environment. NVIDIA ACE provides the underlying suite of models and engineering tools that the feature uses, enabling the different components to run together within the game.

How does PUBG Ally work?

PUBG Ally’s runtime pipeline couples automatic speech recognition, a 2B-parameter small language model, and text-to-speech to interpret player input and generate spoken responses. Voice input from the player is handled by automatic speech recognition, the transcribed intent is processed by the 2B-parameter small language model, and output speech is produced with text-to-speech.

Those three technical pieces form the interactive loop that lets the AI teammate understand players and speak back. NVIDIA ACE supplies the efficient models and tooling used across this pipeline, so each stage—ASR, the small language model, and TTS—runs as part of the ACE-powered system inside PUBG: BATTLEGROUNDS.

Why it matters

PUBG Ally replaces the scripted behavior trees and fixed dialogue that have long constrained in-game companions with a model-driven pipeline that accepts spoken input and returns generated speech. Moving from fixed dialogue to an ASR-to-LM-to-TTS flow changes what players can expect from an AI teammate: more flexible, context-aware responses instead of pre-authored lines.

The use of a 2B-parameter small language model signals a design tradeoff toward efficiency and on-device or low-latency inference, rather than relying on very large models. That choice reflects NVIDIA ACE’s emphasis on efficient models and tooling and suggests a path for other game teams seeking conversational companions without the resource cost of massive models.

What to watch

Whether KRAFTON expands PUBG Ally’s role within PUBG: BATTLEGROUNDS or adapts the same ACE-powered pipeline to other game modes will be the immediate follow-up to monitor. Pay attention to how the team adjusts the 2B-parameter model and the ACE tooling for scale, latency, and multiplayer interaction, since those are the levers that determine real-time in-game performance.