Production voice clone (PVC)
A production voice clone is a higher-fidelity synthesized voice produced from a longer training session (typically 5-15 minutes of varied audio). PVC handles edge cases (uncommon names, dramatic emotional range, technical terms) better than instant voice clones.
Why it matters
Production voice cloning is the higher-effort, higher-quality cousin of instant voice cloning (IVC). Where IVC needs ~30 seconds and produces a clone good enough for most calls, PVC needs 5-15 minutes of varied audio (formal, conversational, list-reading, emotional range) and produces a clone that is meaningfully better on edge cases.
When PVC matters:
- Uncommon proper nouns. The IVC may stumble on a customer's last name; PVC handles it more naturally.
- Technical terms. Industry-specific vocabulary (HVAC parts, medical terms, legal jargon) is more accurate on PVC.
- Dramatic emotional range. PVC handles "I'm sorry to hear that" with more believable warmth.
- Long monologues. PVC keeps cadence over 60-second responses better than IVC.
Most service-business calls do not hit these edges. The IVC is sufficient. PVC is the upgrade path when an owner notices quality issues, not the default.
The infrastructure trade-off: PVC takes longer to produce (a real recording session, not a wizard prompt) and uses more compute. The cost difference is real but small at typical call volumes.
How Ajeris handles this
Ajeris's business agent defaults to IVC for fast setup. Owners who want PVC can upgrade through the manage page; the agent walks through a 10-15 minute recording session reading a curated prompt set designed to capture the full cadence range.
Once trained, the PVC voice replaces the IVC voice for all subsequent calls. The owner can switch back if needed. Both clones live in the user's encrypted store; only the owner can use them.
