Using the Lorvox AI platform

Lorvox AI analyzes 100% of your customer calls. Each recording moves through a fixed pipeline:

upload / integration sync
   ↓  queued
   ↓  transcribing      Deepgram, language detection + diarization
   ↓  transcribed       timestamped sentences, speaker labels
   ↓  processing        PII redaction, then AI analysis
   └→ analyzed          sentiment, topics, QA score, compliance flags

• · Transcription — word- and sentence-level timings, speaker separation, automatic language detection across 30+ languages and 17+ Arabic dialects.
• · Analytics — sentiment, topics, talk ratio, handle time, agent QA scores.
• · Compliance — greeting, identity verification and disclosure checks with risk flags.
• · Delivery — dashboard, per-call and bulk exports, and a REST API for your data warehouse.

1. Receive your invitation, then sign in at /auth. Forgot your password? Use Forgot password? to get a reset email.
2. Confirm your workspace under Dashboard → Company → Profile — name, timezone, default language and recording retention.
3. Invite your team under Company → Team and assign roles.
4. Upload a first call from Dashboard → Calls, or connect a contact center platform under Dashboard → Integrations.
5. Review results on Dashboard → Analytics once the call reaches the analyzed status.

Open Dashboard → Calls and use the upload panel. Each upload stores the audio in your company's private recording folder and immediately queues transcription.

• · Supported audio — common formats such as MP3, WAV, M4A, OGG, FLAC and WebM.
• · Optional metadata — agent name and customer phone, used to group agent scorecards.
• · Language — leave on auto-detect or pin a specific language/dialect (see below).

Status moves uploaded → queued → transcribing → transcribed → processing → analyzed. A failed call shows the error inline and can be retried from the row actions.

Drop up to 50 files per batch into the upload area. Each file is uploaded, transcribed and analyzed automatically — no per-file clicks.

• · The queue panel shows per-file progress, current stage and errors.
• · Batch settings (language, agent name) apply to every file in the batch; you can edit individual calls afterwards.
• · Processing runs in the background — you can close the tab and come back later.
• · Failed files stay in the list so you can retry only those.

Every call can be transcribed with automatic language detection or with a pinned language tag. Auto-detect is the default and writes the result to the detected_language field shown in the calls table.

• · Auto-detect — best when a queue mixes languages; the engine identifies the language before transcribing.
• · Pinned language — best accuracy when you know the call language. Includes 17+ Arabic dialects (Egyptian, Gulf, Levantine, Maghrebi), English variants (US, UK, AU, IN), Spanish, French, Portuguese, German, Hindi, Chinese variants and more.
• · Multi-language — for calls that code-switch between two languages in the same conversation.

Set a workspace default under Company → Profile → Default language so uploads and integrations inherit it.

Open any call row to expand the transcript view. Each sentence carries a timestamp and a speaker label, so you can jump straight to the moment that matters.

[00:00] Agent    Thank you for calling, my name is Sara. How can I help?
[00:07] Customer  I was charged twice for my subscription this month.
[00:14] Agent    I'm sorry about that — may I verify your account first?

• · Speaker separation — diarization splits the conversation into speakers; the dominant opening speaker is labelled as the agent where it can be inferred.
• · Timings — sentence-level start/end times are stored in utterances and available through the API.
• · Summary — an AI summary and key topics sit above the transcript.

Dashboard → Analytics aggregates every analyzed call for your company. Metrics are computed per call and rolled up into daily snapshots.

Per-call metrics

• · Sentiment — a label (positive / neutral / negative) plus a score from -1 to 1.
• · CSAT proxy — predicted satisfaction derived from tone, resolution and closing behaviour.
• · QA score — 0–100 agent score across greeting, discovery, empathy, resolution and closing.
• · Talk ratio & silence ratio — share of speech time per side and dead air.
• · Interruptions — count of overlapping speech events.
• · Handle & hold time — total handling duration and time on hold.
• · Topics — ranked topics detected in the conversation.

Workspace roll-ups

• · Total calls and minutes, and how many have been analyzed.
• · Average CSAT, sentiment split, and sentiment trend over time.
• · Average QA score and top agents by score.
• · Calls by source (upload vs. each connected platform).
• · Compliance adherence percentage and flagged-call count.

Snapshots refresh automatically; use Refresh on the analytics page to recompute on demand after a large batch.

Each analyzed call is checked against a standard rule set:

• · Greeting — agent identified themselves and the company.
• · Verification — customer identity verified before account changes.
• · Disclosure — mandatory statements (recording, terms, fees) were delivered.

The result is a compliance_score, a set of risk_flags, and a flagged boolean for calls that need human review. Flagged calls surface at the top of the analytics page and in the API.

From Dashboard → Calls you can export in three ways:

• · Per call — use the row actions menu for TXT (formatted [MM:SS] Speaker: text), CSV (one row per utterance) or JSON (full transcript, timings and analytics).
• · Bulk ZIP — export the current filtered list and choose which formats to include (TXT, CSV, JSON, or all) to keep archives small.
• · REST API — for scheduled syncs into a data warehouse. See the API reference.

Dashboard → Integrations connects Lorvox to your contact center so recordings arrive automatically. Genesys Cloud is fully supported, with adapters for Five9, NICE, Talkdesk, Twilio, Aircall, RingCentral, 8x8, Zoom CC and AWS Connect.

1. Pick your provider and enter its credentials (OAuth client or API key).
2. Save — Lorvox validates the credentials and stores them encrypted, then reports connected.
3. New recordings are discovered, downloaded, checksummed, transcribed and analyzed automatically.
4. Optionally run a backfill over a date range to import history.

The integration detail view shows sync state, last sync time, recent errors and a request log for troubleshooting.

• · Viewer — read calls, transcripts and analytics.
• · Analyst — viewer, plus upload calls and run exports.
• · Admin — manages the company: team members, integrations, API keys, billing and retention.
• · Super admin — Lorvox staff only; manages companies, users and platform health.

All data is isolated per company. Every query is scoped to your workspace at the database level, and admin actions are recorded in Company → Audit log.

• · Dashboard → Billing shows your plan, included minutes, current usage and invoices.
• · Minutes are counted from analyzed call duration; usage beyond the plan is billed at the overage rate on your plan.
• · Retention — choose to keep recordings or discard them after processing under Company → Profile. Transcripts and analytics are always kept for the configured retention window.