Call Transcription & AI Summarization

Short answer: SIP.IO turns a finished call recording into searchable text and an AI summary, sentiment, key points, action items, and disposition, automatically or on demand. Use the built-in engine, billed per minute of audio, or bring your own OpenAI-standard AI provider’s API key if you want to.

What you can do

• Transcribe automatically. Turn on account-wide auto-transcription and every new recording gets a transcript without any extra API call.
• Transcribe on demand. Call the API for a specific recording, or batch-transcribe a set of recordings by id or date range.
• Get an AI summary, not just text. Sentiment, key points, action items, and a call disposition, generated from the transcript.
• Search across calls. List and filter transcripts, not just individual recordings.
• Bring your own AI provider. Use SIP.IO’s built-in engine, or connect your own API key for any OpenAI-standard transcription or LLM endpoint.

How it works

Transcription runs on the finished call recording: once a recording’s audio is available, you can transcribe it, so recording needs to be enabled for the calls you want transcribed.

# transcribe a specific recording, and summarize it in the same call
curl -X POST https://api.sip.io/v1/recordings/rec_.../transcribe \
  -H 'x-api-key: sk_…' -H 'content-type: application/json' \
  -d '{ "summarize": true }'

# read it back
curl https://api.sip.io/v1/recordings/rec_.../transcript -H 'x-api-key: sk_…'
# → { text, language, summary, summary_data: { sentiment, key_points, action_items, disposition, topics } }

Prefer it automatic? Set your account’s auto-transcribe (and auto-summarize) defaults once, and every new recording is transcribed without calling the API per call. Need to catch up on older recordings? POST /v1/transcripts/batch accepts a list of recording ids or a date range and works through them in the background.

Built-in engine or your own provider

SIP.IO’s built-in transcription and summarization is ready to use with no setup, billed per minute of audio transcribed (see Pricing). If you’d rather use your own AI provider, for a specific vendor relationship, data-handling requirement, or model preference, connect your own API key for any OpenAI-standard transcription or LLM endpoint, and SIP.IO routes transcription through it instead:

curl -X PUT https://api.sip.io/v1/transcription/config \
  -H 'x-api-key: sk_…' -H 'content-type: application/json' \
  -d '{ "provider": "openai-compatible", "api_key": "sk-…", "base_url": "https://your-provider.example.com/v1" }'

Your key is encrypted at rest. With your own provider connected, that provider bills you directly for its own usage, and SIP.IO adds a small per-minute platform fee on top, shown alongside the rest of your usage in /v1/usage and your monthly statement.

Retention: 30 days free, up to 7 years

Every recording and every transcript is stored free for 30 days. Need it longer, for compliance, QA, or dispute resolution? Extend retention on a per-call or account-wide basis, up to 7 years, for a small storage add-on billed only on what you keep past the free window. See Pricing for the current rate.

Capability table

FAQ

How does call transcription work? Once a call recording finishes processing, transcribe it automatically or on demand, then generate an AI summary. Recording needs to be enabled for the call.

Does it cost extra? Yes, the built-in engine is billed per minute of audio transcribed. Bringing your own AI provider adds a small SIP.IO per-minute platform fee on top of what that provider charges you directly.

Can I use my own AI provider? Yes, connect your own API key for any OpenAI-standard transcription or LLM endpoint and SIP.IO routes transcription through it.

How long are transcripts and recordings kept? 30 days free, extendable up to 7 years for a small storage add-on. See Pricing.

Start free · Call Recording · Pricing