Call Intelligence
SalesHunter AI — enter password to continue
Call Intelligence
Mode
Choose whether the tool calls AI automatically, or transcribes free in your browser. Both modes now record and boost audio the same way, so both pick up a quieter speakerphone voice. Only Full auto tells the agent and client apart — that needs AssemblyAI's diarization; the free local model transcribes accurately but doesn't separate speakers.
Leave on Auto-detect unless a call consistently gets mis-transcribed — forcing a language on quiet audio can occasionally cause repeated-phrase glitches.
Upload a call recording
Click to upload or drag an audio file here
MP3, WAV, M4A, or exported Zoom/Meet recording — any length, large files are auto-chunked
Prefer to paste a transcript instead? →
Record a live call
Works for calls on speaker near your mic. For a cleaner result on Zoom/Meet, use "Upload" with the platform's own recording instead.
Keep this tab open and in the foreground — phones pause background tabs' mic access, which cuts recordings short.
Use a second device. If you're taking the call on your phone, run this tool on a different phone, tablet, or laptop sitting nearby with the call on speaker. Most mobile browsers can't access the microphone at all while that same device is on an active phone call — that's an OS-level restriction, not something a webpage can override.
Recording is boosted automatically to pick up a quieter speakerphone voice — place the phone's speaker as close to the mic as you can. No-API mode transcribes locally in your browser (first use downloads a small speech model, ~140MB) — accurate, but it can't separate the agent's voice from the client's the way Full auto's AssemblyAI does.
Past calls
Calls are automatically deleted 2 days after the call date.
Tools used & how it works
AssemblyAI — transcription + speaker separation
Turns the recording into text and tells the agent's voice apart from the client's. Get a free key at assemblyai.com, then paste it into sh-proxy.php on your server (ASSEMBLYAI_API_KEY). Needed for Upload and Record in Full auto mode.
Anthropic Claude — the client brief
Reads the transcript and writes the BANT-style summary. Get a key at console.anthropic.com, add it to sh-proxy.php (ANTHROPIC_API_KEY).
sh-proxy.php — your own server
Upload it anywhere on your WordPress/IONOS hosting, then paste its URL into Settings → Proxy URL. Keeps both API keys off the browser.
Full auto vs. No-API mode
Full auto (needs the keys above): AssemblyAI transcribes and separates agent from client; Claude writes the brief. No-API: free — transcribes locally in your browser using a small Whisper model (no key, no server), then an on-device keyword extractor builds the brief. Accurate transcription either way; only Full auto tells the two speakers apart.
Using it
Upload a recording, or Record live (keep the tab open and in the foreground). The brief appears automatically — click any call to expand the full write-up, edit fields inline, or delete it. Filter History by date or search.
Recording during a live call
If you're taking the call on your phone, run this tool on a different device nearby with the call on speaker. Most mobile browsers block microphone access entirely while that same device is on an active phone call — a platform restriction, not a bug in this tool.
Sharpen the summary with Claude.ai (optional)
1. Copy the prompt below → 2. paste into a Claude.ai chat → 3. paste Claude's reply back here.