Voice brain-dump app — talk, stop, get structured notes. Runs on Mark's server in Georgia. No data leaves to Google or OpenAI.
Access
bidetai.thebarnetts.info
6910
PIN — share only with people you trust
Basic flow
1
Hit RecordTalk for as long as you want. 2 minutes to 45+. Don't worry about pauses, rambling, or "ums" — it handles it.
2
Hit StopGPU transcription starts automatically. Large-v3-turbo Whisper model. Usually done in 30–60 seconds for a 10-minute dump.
3
Wait ~30–60 more secondsAI analysis runs on the transcript. Gemma 12B local model, no cloud fees.
4
Get your outputFour panels ready to use.
What you get
Raw Transcript
Word-for-word what you said. Whisper GPU output, unedited.
Clean Transcript
Filler words removed, sentences formed. Readable version of what you said.
Analysis
Decisions, Priority Signals, Open Questions, Action Items — extracted by the AI.
For-AI Paste
Formatted output optimized to hand directly to Claude or any AI for follow-up. Paste it into a chat, get instant context.
Class Mode
Class Mode — toggle ON before recording
No auto-stop timer. Auto-saves draft every 30 seconds. Good for long sessions where you don't want it to cut off.
Past sessions
Tap the clock icon to see previous dumps. Clean and raw transcripts are always there. Full AI output panel coming soon.
FAQ
Is this free to use?
Yes. It runs on Mark's Apex server. Transcription uses GPU (no cloud cost). AI analysis uses a local model (no cloud cost). The only cost is the server electricity, which is already running 24/7.
How long can I record?
No hard limit. 45 minutes works fine. Beyond that, analysis takes longer but still works. Short dumps (under 2 minutes) may have thin AI output — more words = better analysis.
What if it's slow or stuck?
The server is in Georgia and occasionally restarts. If a job spins for more than 3 minutes, refresh the page and try again. The stuck job clears automatically on next restart. Text Mark if it's consistently broken.
Is my data private?
Yes. Audio is processed on-device (GPU on the Apex server in Mark's home). Transcripts are stored on the same server. Nothing goes to Google, OpenAI, or any cloud AI. The PIN is the only access gate.
Can I use it on my phone?
Yes — any browser works. On mobile, use Chrome or Safari. Tap the microphone button, allow mic access when prompted. The site is not a native app (no install required).
What's the For-AI Paste for?
Copy the entire For-AI output and paste it into a Claude, ChatGPT, or Gemini conversation. The format is optimized so the AI gets full context of what you said and can answer questions, write follow-ups, or take action immediately.
This is Mark's personal server. If you abuse it (running it 24/7, sharing the PIN publicly), he'll pull it. Treat it like you're a guest in his house.