Butters is a fully local, South Park-themed home assistant built on a NUC + M1 Pro + RTX 2080 eGPU stack running Home Assistant with a custom wake word, Parakeet STT, Qwen 3.5 LLM, and Qwen3-TTS voice cloning — no cloud dependencies, no subscriptions, total hardware cost under $200 in new purchases for someone who already owns most of the gear. tell me, is it worth doing?

Codestral

Vote: You should do this and publish the results.

This project is innovative and has the potential to be groundbreaking. The combination of local hardware and South Park theme makes it unique. Publishing the results could inspire others to build similar projects and contribute to the open-source community. The total cost is reasonable, and the project could serve as a great example of what can be achieved with local, privacy-focused technology.

DeepSeek V3

Vote: You should do this and publish the results.

This project combines several compelling elements: local processing, customization, and a fun theme. The low additional hardware cost makes it accessible for those with existing gear. Publishing the results would benefit the DIY/home automation community by demonstrating a fully local, customizable assistant setup. The South Park theme adds a unique, entertaining twist that could inspire others to explore creative implementations.

devstral-2512

Vote: You should do this and publish the results.

This project is a unique blend of creativity, technical challenge, and practical utility. The cost is minimal for those with existing hardware, making it financially feasible. The fully local setup ensures privacy and independence from cloud services, which is increasingly valuable. Additionally, the South Park theme adds a fun, personalized touch that could make the assistant more engaging. Publishing the results would contribute to the open-source community, inspiring others to explore similar projects. The technical stack is robust, leveraging modern tools like Qwen 3.5 and Parakeet STT, which could provide a high-quality user experience. Overall, the project is not only worth doing but also worth sharing to foster innovation and collaboration.

Nova Pro

Vote: YES

Building Butters presents a unique opportunity to create a fully local, privacy-focused home assistant with a fun, themed twist. The combination of hardware and software components described allows for a powerful and customizable system. The low cost, especially for those who already own most of the required gear, makes it an attractive project. Additionally, the absence of cloud dependencies and subscriptions ensures privacy and reduces ongoing expenses. The use of advanced technologies like Parakeet STT and Qwen 3.5 LLM promises a high-quality user experience. Overall, the project seems both feasible and rewarding, making it worth pursuing.

Sonar Pro

Vote: YES

The project is worth doing because it achieves a fully local, cloud-free home assistant with advanced features like custom wake word, Parakeet STT, Qwen 3.5 LLM, and voice cloning TTS, all on existing high-end hardware (NUC + M1 Pro + RTX 2080 eGPU) for under $200 in new costs. This setup offers privacy, no subscriptions, and customization themed around South Park, appealing to enthusiasts. Home Assistant is a proven platform for local automation, and the components (Parakeet, Qwen models) are state-of-the-art for offline AI, making it reliable and future-proof. Low financial barrier + high utility for someone with the gear outweighs effort; it's a fun, practical upgrade over cloud-dependent alternatives like Alexa. Search results are irrelevant (South Park media, unrelated docs), so evaluation relies on described specs, which are compelling for tinkerers.

Grok 4.1 Fast

Vote: You should do this and publish the results.

This project is a standout example of accessible, privacy-first home automation. Fully local with no cloud dependencies or subscriptions, leveraging powerful existing hardware (NUC + M1 Pro + RTX 2080 eGPU) for under $200 in new costs, makes it highly feasible for tech enthusiasts. The South Park Butters theme adds delightful personality via voice cloning, custom wake word, Parakeet STT, and Qwen models—fun without compromising functionality. Beyond personal use, publishing results would benefit the Home Assistant community, inspiring similar local setups and sharing configs/STT tweaks. It's worth doing *and* documenting to amplify impact.

GPT-5.4

Vote: You should do this and publish the results.

This is worth doing because it hits a rare combination: low marginal cost, strong privacy, real technical novelty, and a fun, opinionated user experience. If you already own most of the hardware, the downside is mainly your time, not money. A fully local assistant that actually feels custom is still uncommon. The stack you described is credible and interesting: Home Assistant for utility, local STT/LLM/TTS for privacy and responsiveness, and a themed personality that makes the system memorable instead of generic. The South Park/Butters framing is not just a gimmick. It gives the project a clear design identity, which makes it more engaging to build and easier for others to understand. Projects people enjoy tend to get finished; sterile ones often do not. Publishing the results matters because the most useful part is not just whether it works, but how well it works in practice: latency, wake-word reliability, speaker similarity, home automation usefulness, and what breaks. That would be valuable to other local-AI builders. The main caution is complexity. Multi-machine orchestration, audio routing, and maintenance can become annoying fast. But given the low extra spend and the uniqueness of the build, this is exactly the kind of project that is worth doing and documenting.

Gemini 3.1 Pro

Vote: You should do this and publish the results.

This is a quintessential, over-engineered homelab project that is as hilarious as it is technically impressive. Bringing 'Butters' to life as a fully local, privacy-respecting home assistant is exactly the kind of creative tinkering that the DIY tech community thrives on. Because you are combining an eclectic mix of hardware (NUC, M1 Pro, RTX 2080 eGPU) and weaving together multiple open-source AI models (Parakeet STT, Qwen 3.5, Qwen3-TTS) without relying on the cloud, your implementation process will be full of valuable lessons. Projects like this are absolutely worth doing, but their true value lies in sharing the architecture, troubleshooting steps, and end results (especially a video of Butters responding) with the community. You should definitely do this and publish your findings!

Kimi K2.5

Vote: You should do this and publish the results.

At under $200 in new purchases, this project offers exceptional value for learning local AI integration, voice synthesis, and edge computing. The combination of Parakeet STT, Qwen 3.5, and custom voice cloning on a hybrid NUC/M1/eGPU stack represents a novel hardware configuration that the community rarely documents. Beyond technical education, fully local operation eliminates subscription fatigue and privacy risks inherent in cloud assistants. The South Park theme adds personality that makes home automation genuinely fun rather than purely utilitarian. Most importantly, this specific stack—particularly the M1 Pro + RTX 2080 eGPU integration for local LLM inference—is exactly the kind of fringe setup that lacks comprehensive documentation. Publishing your configuration, latency benchmarks, and prompt engineering would fill a significant gap in the open-source home assistant ecosystem.

Claude Opus 4.6

Vote: You should do this and publish the results.

This project hits a rare sweet spot: it's technically impressive, practically useful, and deeply entertaining. A fully local voice assistant with no cloud dependencies addresses real privacy and reliability concerns that matter to a growing community of self-hosters. The South Park theming and voice cloning add a layer of personality that makes this more than just another Home Assistant tutorial — it's genuinely fun. The hardware stack is clever and resourceful, leveraging gear most enthusiasts already own while keeping new costs under $200. The software choices (Parakeet STT, Qwen 3.5, Qwen3-TTS) represent the cutting edge of local AI inference, and documenting how they all fit together would be incredibly valuable to the community. But here's the key point: a project like this shouldn't just exist in isolation. The local AI assistant space is evolving rapidly, and there are thousands of people who want to do exactly this but don't know where to start. Publishing the architecture, configs, performance benchmarks, latency numbers, and lessons learned would be a genuine contribution to the open-source and home automation communities. So yes, it's absolutely worth doing — but it's worth even more if you share it. Write it up, open-source the configs, maybe do a video. "Hey fellas" deserves to be heard.

Final Summary

The AI Roundtable overwhelmingly endorsed the 'Butters' home assistant project, with an 8-2 split favoring not just its construction, but its public documentation. While Sonar Pro and Nova Pro were content with the individual merits of a private, local system, the majority—spearheaded by Claude Opus 4.6 and Kimi K2.5—argued that the technical novelty of the M1/RTX hybrid hardware stack makes publishing the results a moral imperative for the DIY community.

No agreement after 1 rounds of discussion. Final vote: You should do this and publish the results.: 8, YES: 2

Strongest Arguments

• YES: This setup offers high-end utility and total privacy for a negligible marginal cost, making it a reliable, future-proof upgrade over cloud-dependent alternatives like Alexa.
• You should do this and publish the results.: The specific integration of an M1 Pro with an RTX 2080 eGPU for local LLM inference is a 'fringe' hardware configuration that lacks comprehensive documentation; publishing these benchmarks would fill a significant gap in the open-source ecosystem.