Cloning gives you a voice in 60 seconds. Training gives you a voice that's been polished — calibrated against your brand's pace, energy, and stylistic quirks. The difference matters when you're committing to a voice that'll narrate 500 episodes, 10,000 lessons, or every product video for the next two years.
What you get
- Available on Creator, Pro, Scale, Business and Enterprise plans
- Project-tracked: every training job has its own audit record
- Consent script required at submission — built-in safety
- Trained voice lands directly in your Library on completion
- Reusable across all project types
- Per-team training history (admins can review usage)
Why train instead of clone?
Voice cloning is fast and good enough for most short-form content. Training takes longer but gives you something closer to a custom-built brand asset: stronger consistency over long runs, better handling of edge-case prosody, and finer control over emotional range. If you're producing serialized content (podcasts, courses, audiobook series) — train.
Sample preparation matters.
The best training results come from 3–10 minutes of varied, clean audio recorded in a single environment with a single mic. Avoid music, reverb, and crosstalk. Speak naturally — read a varied passage that includes statements, questions, exclamations, and a few longer narrative arcs. The model needs to see your full range.
Per-project tracking, not a black box.
Every training job is attached to the project that owns it. Your team can see who started the training, when it ran, and which voice it produced. For agencies and enterprises this audit trail is non-negotiable — for solo creators it just keeps the workspace tidy.
Quick answers
How long does training take?
Typically 5–30 minutes depending on sample length and provider load. You'll get an email when it's done.
What's the consent script for?
It's a short statement you read out loud confirming you own the voice or have written permission. We keep it on file in case the voice is ever flagged.