psmon
177 posts


@rizamarhaban “Wow, it looks like a World of Warcraft-style world conquest simulation game!”
English

AgentZero
Supertonic is an ultra-fast on-device TTS engine with 10 built-in voices and support for 31 languages including Korean.
It runs locally with ONNX models, delivers real-time-factor speeds as low as ~0.012 on an M4 Pro CPU, and integrates with AgentZero through a pip-installed Python subprocess workflow.
The setup includes smart multi-Python detection, live download progress, actionable error diagnostics, and an easy-to-use voice settings interface.
SuperTonic:
github.com/supertone-inc/…
AgentZero:
#models" target="_blank" rel="nofollow noopener">psmon.github.io/AgentZeroLite/…


English

@Agredo10 We are planning to onboard an ONNX-based voice model for our upcoming platform evaluation.
Supertonic by Supertone is currently one of the top candidates due to its fast local inference, lightweight deployment, and offline ONNX Runtime support.
GitHub:
github.com/supertone-inc/…
English

Running Gemma 4 On-Device in .NET
A Journey of Failure and Success
⸻
🎯 What We Wanted
“On a user’s PC, without internet, without a separate server like Ollama,
we want to directly load Gemma 4 in a .NET app and generate tokens.”
more:
github.com/psmon/AgentZer…

English

If you're a developer, it takes just 3 lines.
pip install supertonic
from supertonic import TTS
tts = TTS(auto_download=True)
wav, _ = tts.synthesize(
text="안녕하세요",
lang="ko",
voice_style=style
)
License: MIT for the code, OpenRAIL-M for the model.
Commercial use? Yes — as long as you comply with the responsible-use terms.
Sources
Supertonic GitHub repository: github.com/supertone-inc/… (accessed 2026-05-16)
Version 3 update (2026-04-29): added support for 31 languages and 10 expressive style tags
Supertone Inc. headquarters: Gangnam-gu, Seoul (HYBE subsidiary, CEO Hoon Heo) — supertone.ai/en/company
Benchmark comparison: README “Natural Text Handling” section (vs. ElevenLabs Flash v2.5 / OpenAI TTS-1 / Gemini 2.5 Flash TTS)
Model licensing: MIT for code, OpenRAIL-M for model, PyTorch BSD 3-Clause for training

English
psmon retweetledi

A great question in the Netclaw Discord: "what should I use this for?"
I'm a business owner and have dozens of use cases that apply to me professionally; I'll share those AND some personal use cases.
Aaron Stannard@Aaronontheweb
So I haven't made a YouTube video or done like a thousand other things I wanted to before I made this public, but since I've already made it OSS... Netclaw is available and ready for use. Netclaw == Simple, secure, reliable agents. Open source. Built with .NET. Local inference
English

“Currently experimenting with a variant that represents Akka.NET code as outer space…”
next : github.com/psmon/CodeScan

English

Building LLM Agents in Akka.NET — Porting Akka.io's Agent SDK to the Actor Model
luxuriant-brazil-09c.notion.site/Building-LLM-A…

English

CodeScan
Another variant project created
Three interfaces coexist
CLI: A command interface where the AI terminal handles everything automaticallySupports full CRUD operations
TUI: A terminal user interface for cases where human intervention and editing are still neededOccasional updates
GUI: A monitoring interface for when ASCII text becomes hard to read due to eye strainRead-only mode
github.com/psmon/CodeScan



English

@Aaronontheweb Just borrowing a few concepts. Still hoping for true OSS-spirit Akka.NET updates someday too.
English
psmon retweetledi

有人把 GPT-Image-2 最强提示词 全偷出来了!
GitHub这个仓库直接整合576个高质量实测案例,人像写真、海报设计、角色概念、UI Mockup全覆盖!
从便利店霓虹灯风、赛博朋克到宋朝社交媒体界面……
全是真实创作者跑出来的神级Prompt,直接抄作业就行!
不会写提示词的兄弟,这波直接救命!
github.com/EvoLinkAI/awes…

中文

🚀 AgentZero AICLI
AI for Everyone — Including the Visually Impaired 👨🦯✨
🎙️ Voice-first AI CLI
🛒 Search & shop with natural conversation
🧠 Smart multi-agent workflow
💻 Accessible, keyboard-friendly design
#dashboard" target="_blank" rel="nofollow noopener">psmon.github.io/AgentZeroLite/…
Technology should have no barriers.
#AI #Accessibility #AICLI #AgentZero #akka.net

English
psmon retweetledi

发现个 Claude Code 的神级 Skill,画图颜值真的吊打 Mermaid 和 Draw.io!
fireworks-tech-graph,只要你一句话,它直接给你吐出生产级别的 SVG+PNG 技术图。
最绝的是它有 7 种神仙画风:深色终端风、工程蓝图风、毛玻璃风,甚至还能模仿 OpenAI 和 Claude 的官方配色风格。
支持 14 种 UML 图,像 RAG 架构、多智能体协作这种复杂图,一句"帮我画个 RAG 架构图,深色风格"直接搞定。
安装贼简单,一行命令 `npx skills add yizhiyanhua-ai/fireworks-tech-graph` 就能装到 Claude Code 里,输出 SVG 方便二次编辑,高清 PNG 直接贴文档。
平时写技术文档、做汇报 PPT 的,这个必须试一下。
项目地址放评论区了👇
中文
psmon retweetledi
psmon retweetledi
psmon retweetledi

에이전트 워크플로우 짜는 개발자면 이 글은 키핑해둘만함. Osmani가 AGENTS.md, 훅, 샌드박스, 서브에이전트, 컨텍스트 압축을 한 묶음으로 정리해놨음.
파일시스템 접근, bash, 샌드박스, 테스트 훅, 컨텍스트 압축이 없으면 똑똑한 모델도 중간에 멍청해짐.
Addy Osmani@addyosmani
한국어








