NPCTalk

NPCTalk is a toolkit for AI coding agents to add voice interactions and cloned NPC speech to games.

NPCTalk is a toolkit that lets AI agents mod games by enabling any NPC to talk back in character using Fish Audio voices. It provides skills, a self-hosted voice server, and tools for AI coding agents to wire up voice input, dialogue generation, and in-game actions. The toolkit supports Fish Audio Drama 3 and S2 Pro models hosted on Modal or fal, Cerebras for LLM lines, NVIDIA Nemotron ASR for speech-to-text, and TypeSafe Jev for actions. It includes a complete example for GTA V on Windows, along with evaluation tools to measure voice similarity and word error rates. The project is designed for single-player games or self-hosted servers, using environment variables for credentials and enforcing safety checks.

NPCTalk at a glance

Best for: Software engineer
Runs on: Web, Windows, API, VS Code, Cursor, CLI, self-hosted, open source
Pricing: Free
Released: October 2026
Category: AI agent tool
Website: github.com/christos-antono/npctalk

Key features

Self-hosted Fish Audio voice server
Agent skills for AI coding agents
NPCTalk CLI for health checks and voice generation
Voice likeness evaluation metrics
NPC dialogue generation with personas
In-game action mapping
Voice input with speech-to-text

Use it for

Make in-game NPCs talk back using cloned voices
Set up a Fish Audio voice server on Modal
Build a voice-enabled mod for GTA V story mode
Measure speaker similarity and voice likeness

Alternatives to NPCTalk

pstack — Improves AI coding agent outcomes with rigorous workflows.
NovaSynth — Tests voice agents on real calls and validates fixes automatically.
NemoClaw — Provides managed security, networking, and lifecycle operations for AI agents.
Acontext — Builds human-readable Markdown skill files from agent runs without using embeddings.
SkillClaw — Automatically evolves and unifies agent skills across devices and team members.
Coval — Prove voice agent reliability through simulation and human-in-the-loop testing.
Pipecat — Builds real-time voice and multimodal agents with frame-level control.
Alook — Connects existing local coding agents to team communication channels.
OmniAgent — It evolves skills in real-time and dynamically hardens security during task execution.
Waku Agent — Local-first memory storage and readable Python code for AI agent logic.
Nexent — Auto-generate production-grade AI agents from natural language prompts without complex setup.
VoiceGremlin — Validate AI voice agents with real phone calls and plain English test cases.

All NPCTalk alternatives