Show HN: BYO AI free notetaking with optional screen reading for OpenClaw/hermes (stagewhisper.io)

1 points by piotraleksander 2d ago ↗ HN

I've built fully on-device (macOS) meeting transcription/summaries/action-items app. It uses parakeet and gemma 4 or OpenClaw or Hermes Agent to drive main functionalities. It's free, requires no account or email, and is source available. You can also natively connect your own local/VPS OpenClaw or Hermes to use them, their memory and context, as a notetaker.

About me: my name is Piotr. I'm having plenty of calls as a freelancing programmer, and have interest in audio domain, so it was natural for me to pick this as my side project. I struggled a bit with echo cancellation, and writing idiomatic plugins to let users use their OpenClaw or Hermes agents as BYO AI (basically backend). I started project coding Tauri app by hand, only after establishing architecture I started leaning on AI coding agents. I use Claude Code with https://github.com/openai/codex-plugin-cc for "adversarial-reviews".

On the website I've included an offer to upgrade StageWhisper Lite to Founders edition with one-time payment: this give screen reading (via accessibility tree) and proactive, two-way communication - that is, your connected agent can provide suggestions by listening to audio (Cluely style), and you can chat with it during the call. Again, you bring your own AI, so Founders is a single payment, no subscription offer.

Both apps include MCP and in-app chat with your transcripts and meeting notes. Both also allow to connect any model downloaded from HF or connecting OpenClaw/Hermes.

I created StageWhisper Lite & Founders to test business feasibility of one-time payment AI apps that have no marginal costs (you either use local model or your own assistant).

I hope you will enjoy free Lite app and consider upgrading to Founders.

4 comments

[ 2.7 ms ] story [ 18.8 ms ] thread

piotraleksander 2d ago ↗

feel free to ask me questions here or reach out via "hello at stagewhisper.io"

[dead] baba_yuhei 2d ago ↗

[flagged]

JohnBizBiz 2d ago ↗

Echo cancellation for real-time output is a nasty problem. I hit similar walls with a browser-based take on this (livesuggest.ai) — you get easier distribution but lose a lot of control compared to native. The Founders real-time suggestions piece is what I'd want to try most. What's the end-to-end latency like in practice?

piotraleksander 13h ago ↗

nice, are you using webrtc aec3 or sth custom?

to your question i did a thorough benchmarking (hence late reply) used 18gb m3 pro macpro

full on-device with gemma 4 e2b is ~1.5s since last utterance. vad silence th is set to 600ms

now, when I connect local hermes agent that uses deepseek v4 flash via their api, latency jumps to 3.5s (this of course includes 600s of silence to even trigger pipeline)

btw, suggestions are throttled to one every 20s by default, it's configurable from the app

afaik cluely is waaaaay over 5s, but it makes sense because their users need lots of ai reasoning to answer interview qs :D different target audience then stagewhisper