> ## Documentation Index > Fetch the complete documentation index at: https://docs.pyai.com/llms.txt > Use this file to discover all available pages before exploring further. # Conversation intelligence > Choose Hear, Recap, post-call data capture, or Trace for transcripts, summaries, structured fields, and compliance evidence. Conversation intelligence is not one output. Start with the artifact your product needs: | Need | Product | What it returns | | - | - | - | | Transcript, timestamps, speakers, SRT/VTT | Hear async jobs | Timestamped transcript artifacts | | Summary, action items, talk ratio, call signals | Recap | Retained `recap.record.v1` plus utterances | | Your own JSON fields from an Omni call | Post-call data capture | Agent-specific JSON Schema result to a signed webhook | | Compliance rules, violations, evidence, exposure | Trace | Per-call scorecard and organization views | These products compose, but they do not replace one another. Hear transcribes audio. Recap consumes a completed transcript. Post-call data capture applies an agent-specific schema. Trace evaluates eligible calls against compliance rules. ## Speaker diarization: who said what [Speaker diarization with Hear](/guides/speaker-diarization) groups a recorded conversation into speaker turns with timestamps. Use mono diarization or existing stereo channel separation to prepare transcripts for review. Speaker labels do not verify a person’s identity or determine their business role. ## Fastest path from a recording to notes 1. Mint a sandbox key from the [Quickstart](/quickstart). Sandbox organizations include Recap scopes and start with Recap enabled. 2. Submit a [timestamped Hear job](/guides/async-transcription-jobs) with a stable `call_id`, plus `channel: true` for stereo or `diarize: true` for mono. 3. Poll `GET /v1/recap/calls/{call_id}` until `status` is `complete`. 4. Read `transcript` for speaker-labelled utterances and `record` for the summary, actions, next steps, talk ratio, signals, and fields. Scaffold the complete path: ```bash theme={null} npm create pyai-app@latest recap-call-intelligence ``` The [Recap guide](/guides/recap-call-intelligence) contains every supported input path: async recording, Hear streaming, existing utterances, packs, webhooks, CRM configuration, retention, and troubleshooting. ## Recap vs post-call data capture Use Recap when the object should be consistent across calls and agents: summary, action items, next steps, disposition, talk ratio, signals, and pack-defined fields. Use [post-call data capture](/guides/post-call-extraction) when one Omni agent needs a custom JSON Schema such as order id, appointment date, or follow-up-required. It runs after the call and delivers the result to that agent's signed webhook. If both are enabled, treat them as separate artifacts. Do not make one webhook consumer guess whether a payload is a Recap record or an agent extraction. ## Recap vs Trace Recap describes what happened and what should happen next. Trace evaluates whether an eligible call followed configured rules and preserves evidence for review. A call can produce both. Trace does not turn an arbitrary recording into a compliance scorecard merely because Hear transcribed it. Read [Trace eligibility and setup](/guides/trace-guardrails) before promising compliance coverage for imported or external calls. ## Bring calls from another system For a recording you already control, submit a Hear job directly. For continuous imports from supported phone systems, use an [inbound call integration](/guides/integrations-overview). Imports still require the relevant Recap or Trace features to be enabled; connecting a phone system does not silently enable paid processing. Inputs, packs, webhooks, CRM fields, records, and retention. Recording submission, speaker labels, subtitles, and job lifecycle. Agent-specific JSON Schema extraction after Omni calls. Eligibility, rule packs, scorecards, and exposure. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.