Audio Source

Captures Mic (Agent) & System Audio (Caller) as separate tracks. Ensure "Share system audio" is enabled when selecting the tab/window.

Recording in progress

Record mode 2-Track (Mic + System)

Agent (Microphone)0%

Caller (System Audio)0%

Saved locally in IndexedDB storage.

Transcribe recordings you already have (mp3, m4a, wav, ogg, webm, flac, ...). Select one or more files, then run them through the same transcription and AI note process.

Choose audio files

Click to choose files or drag & drop them here

mp3 • m4a • wav • ogg • webm • flac • more — select one or multiple

Each file becomes its own track in the transcript (labeled by file name).

Whisper Speech Recognition

Whisper ASR Model

Model downloads once to browser cache. No API key or server required. Larger models are more accurate but slower and use more memory.

Preparing... 0%

Call Transcript Workspace

Local AI Co-Pilot

Runs an LLM entirely inside your browser (WebGPU/WASM) to extract notes, action items, or follow-ups.

About local AI — what it's best for and its limits

Local AI (In-Browser)

What it does best

Fast, private, free note generation on short-to-medium calls (up to roughly 25k chars / ≈15 min).
Calls containing sensitive information that must never leave your device.
Offline use after the one-time model download.
Quick summaries, action items, and follow-up drafts on typical support/customer calls.

Limitations

Only small models fit in the browser (≤1.5B); reasoning and formatting are noticeably weaker than cloud LLMs.
Memory ceiling: WebAssembly is 32-bit (≈4 GB heap cap). Long calls can hit "tensor too large" / out-of-memory errors; the app then retries with a smaller model or CPU.
Speed depends on the GPU: without WebGPU you get CPU ("WASM") mode, which is many times slower.
Keep transcripts compacted/trimmed — the app auto-compresses them (~45%), which helps a lot.

Need your AI off CPU mode? Pick a topic:

Cloud API — Privacy & Free Key

What gets sent, and where

Cloud mode sends your transcript, prompt instruction, and format reference to the provider you select (OpenAI, Google, Anthropic, etc.).
Requests go directly from your browser to that provider — nothing passes through this site or any other middleman.
Your API key is stored only in this browser (local storage) and is sent only to the provider you configure.
Local (In-Browser) mode remains the fully private option — nothing ever leaves your device there.

Want a free API key? (Google Gemini)

Go to aistudio.google.com/api-keys and sign in with a Google account.
Click Create API Key and name it StaticRecorderPro.
Copy the key somewhere safe (a password manager or note) — you may not be able to view it again later.
Come back here, choose Google Gemini as the provider.
Set the model to gemini-3.1-flash-lite — it works very well for this purpose and gets the most tokens because it's an older model in the family.
Paste your copied key into the API Key field, then press Generate with AI.

Tip: Gemini's free tier has generous limits, so it's the easiest way to try Cloud mode without paying. If you run out of tokens, switch to Local mode or use the Copy Prompt button to paste into any AI chat you already have.

AI Model

Generation Style

Heads-up: cloud mode sends your transcript & prompt to the provider. — Need a free API key?

Provider

Model

API Key

Base URL

The key is saved only in this browser (localStorage) and sent directly from your browser to the provider. Use a restricted/limited key — anyone with access to this device can read it.

Opens chatgpt.com in a new tab as a temporary chat with this prompt (instructions + reference + transcript) pre-filled. Press Enter there to send — a static page can't click Send for you.

AI Prompt Instruction

📝 Bullet Summary

✅ Action Items

✉️ Follow-up Email

💡 Key Decisions

Add AI Prompt

Button Name Prompt Instruction

Format Reference / Template (Optional)

Preparing AI... 0%

AI Generated Insight