# nemotron-halloween

DevFeed: [nemotron-halloween](<https://devfeed.tech/articles/nemotron-halloween-109447.md>)

Original publisher: [Read original article](<https://huggingface.co/blog/nvidia/nemotron-halloween>)

Author: Rick Grandy; Maryam Motamedi

Published: 2026-10-30T17:57:24Z

Content type: article

Language: en

Sources: [Hugging Face](<https://devfeed.tech/sources/hugging-face-blog.md>)

Topics: [Nemotron](<https://devfeed.tech/topics/nemotron.md>), [AI Chat](<https://devfeed.tech/topics/ai-chat.md>), [3D](<https://devfeed.tech/topics/3d.md>), [Self Hosted AI Tools](<https://devfeed.tech/topics/self-hosted-ai-tools.md>), [RTX PRO](<https://devfeed.tech/topics/rtx-pro.md>), [Streaming](<https://devfeed.tech/topics/streaming.md>), [AI Context Management](<https://devfeed.tech/topics/ai-context-management.md>), [real-time rendering](<https://devfeed.tech/topics/real-time-rendering.md>), [Tutorial](<https://devfeed.tech/topics/tutorial.md>)

Tags: [3d](<https://devfeed.tech/tags/3d.md>), [agent](<https://devfeed.tech/tags/agent.md>), [asr](<https://devfeed.tech/tags/asr.md>), [bolt](<https://devfeed.tech/tags/bolt.md>), [halloween](<https://devfeed.tech/tags/halloween.md>), [interactive](<https://devfeed.tech/tags/interactive.md>), [magpie-tts](<https://devfeed.tech/tags/magpie-tts.md>), [nemotron](<https://devfeed.tech/tags/nemotron.md>), [nemotron-3-5-bolt](<https://devfeed.tech/tags/nemotron-3-5-bolt.md>), [projection](<https://devfeed.tech/tags/projection.md>), [tts](<https://devfeed.tech/tags/tts.md>), [workstation](<https://devfeed.tech/tags/workstation.md>)

## AI overview

MonsterHouse is a Halloween projection that listens to visitors, answers as a grumpy pirate ship, and softens when they use good manners. NVIDIA Nemotron Voice Agent coordinates Nemotron ASR, a personality-tuned Nemotron 3.5 Bolt E4B dialogue model, Magpie TTS Zeroshot, and Audio2Face-3D; optional Nemotron 3 Nano Omni describes costumes, while YOLO and ByteTrack track visitors. The system runs across two GPUs in one workstation, with dialogue and voice on one and graphics and animation on the other. Bolt was selected after testing more than a dozen models under local memory and compute limits. Streaming speech and prepared text or cached audio reduce response delay: reported medians are 1.70 seconds for Bolt plus TTS and 0.72 seconds for cached responses, with silence detection adding 300 ms. The author's suggested path is to start with the Voice Agent, manage encounter state in application logic, add prompts or a LoRA for character, then add voice, animation, and optional vision. Real visitor performance and night-time costume recognition remain untested, and the project code is not packaged for release.

## Source excerpt

Or rather... **TL;DR:** MonsterHouse is an interactive projection I created for Halloween. He listens to trick-or-treaters, answers in the voice of a grumpy old pirate ship and softens when visitors remember their manners. He is built on the NVIDIA Nemotron Voice Agent with NVIDIA Nemotron ASR, Nemotron 3.5 Bolt E4B 4GB, Magpie TTS Zeroshot and Audio2Face-3D, and he runs on a single workstation. I wanted the kids visiting my house at Halloween to interact in a fun and unique way.