Voice clone isn't a separate setup screen. It lives as the third option in the
Create your video mode sheet — the same sheet where you pick Voice or Camera (see
Record your voice). Open it and Promoat shows one of three states depending on how much voice you've banked.
1. The recording bank
Every voice take you record in the Voice mode is stored as part of a recording bank tied to your account. You don't manage it directly. It just grows as you record. Cloning needs at least 2 minutes total across all your takes.
TIP
Clone is built from real takes, not extra prompts
Most voice-cloning tools ask you to read filler sentences for calibration. Promoat re-uses the takes you already record for your videos, so the bank fills as you make content. Three to five decent voice takes is usually enough.
2. State A — Gate
If your bank is below 2 minutes, opening Voice Clone shows the gate screen.
Almost there.
Cloning needs at least 2 minutes of your voice.
Record a few Script or Freestyle videos to continue.
Record manually this time
Gate state: title, sub-headline explaining the 2-minute threshold, gradient progress bar, manual fallback CTA.
What's on screen:
A gradient waveform icon at the top.
Sub-headline: "Cloning needs at least 2 minutes of your voice. Record a few Script or Freestyle videos to continue."
A gradient progress bar showing your banked seconds vs the 2-minute target. Below it: "{X}s recorded" on the left, "~{N}m to go" on the right.
A gradient CTA: "Record manually this time". Tapping it sends you back to the mode sheet so you can pick Voice (or Camera) for the current video while you keep building the bank.
HEADS-UP
Script mode and Freestyle mode both bank
The sub-headline mentions Script and Freestyle — the two ways you can use the Voice mic recorder. Scriptmeans you read the teleprompter; Freestyle means you ignore it and improvise. Either way the take is logged to the bank.
3. State B — Ready to clone
Once your bank crosses 2 minutes, the same screen flips to a single action.
Ready to clone.
We'll use your 2-minute recording bank to build your voice model.
Clone Ready: title, sub-headline confirming bank size, single Clone-now button with the credit cost embedded.
Sub-headline: "We'll use your {N}-minute recording bank to build your voice model." {N} is rounded up — three minutes banked reads as "your 3-minute recording bank."
A single gradient button: "Clone now" with a credit badge showing 30 on the right.
What 30 credits buys
30 credits covers the entire clone job: training a voice model and storing it on the provider. After that, every TTS generation from the clone is its own per-character cost (~10 credits per 1,000 characters) — but the model itself is one-time.
Credits deduct when you tap Clone now. If the job fails (almost always due to noisy recordings), the 30 credits are refunded automatically and the screen returns to
Ready to clone
so you can retry.
4. The clone job in flight
Tapping Clone now replaces the button with a live progress label.
Ready to clone.
We'll use your 2-minute recording bank to build your voice model.
Cloning: same screen, but the button is replaced by a spinner + percentage label.
The label updates as the job progresses: "Cloning… 10%" → "Cloning… 20%" and so on. Stages run on the server: queued → fetching recordings → training → completed. The whole thing typically finishes in under 30 seconds.
HEADS-UP
You can leave the screen
The job runs on the server. If you back out of the mode sheet, the clone keeps cooking. When you next open Voice Clone the tab will be in the next state — Has Voice — without any extra confirmation.
5. State C — Has Voice
Once the clone exists, opening Voice Clone shows a different screen every time: a play button for the current take, plus three sliders for tweaking how the voice sounds.
Listen, tweak, then use it.
VOICE SETTINGS
Has Voice: title with cloned badge, audio player, three voice sliders.
The header
Title "Your voice." with a small Cloned badge to its right. Sub-headline: "Listen, tweak, then use it." (or "Ready to generate." if no take has been generated yet).
The player
A row with a circular play button and a Play take label. Plays back a TTS render of the current script through your cloned voice. Tap again to pause.
The voice sliders
Three sliders control how the cloned voice is rendered. Each is a 0.00–1.00 value:
Stability (default 0.00) — Lower values give a more expressive, dynamic delivery; higher values flatten inflection but reduce occasional weird artifacts. Bump up if your clone sounds too erratic.
Similarity (default 0.62) — How tightly the clone tries to match the original recordings. Higher = closer to your real voice but more likely to copy background noise from the bank takes. Lower = cleaner but less identifiable.
Style (default 1.00) — How much of the stylistic delivery (pacing, emphasis) is borrowed from the bank. Lower for neutral readouts, higher for personality.
TIP
The defaults are tuned for short-form video
Most users never touch these. The defaults are calibrated for TikTok / Reels / Shorts narration — slightly expressive, fairly faithful. Move them only if your clone consistently sounds wrong.
Re-cloning
Re-cloning replaces the existing voice with a new one trained on the current bank. Useful if you've recorded better-quality takes since the original clone, or if your voice has changed. Re-cloning costs another 30 credits and triggers the same cloning flow.
6. Using the clone for the current video
Once you've previewed the take and like it, tap Use this take. Promoat captures the rendered audio URL and routes you to the Caption options sheet — same flow as a normal voice recording. The video renders with your cloned voice instead of the default TTS.
HEADS-UP
The clone is per-account, not per-video
Once you have a clone, every Voice Clone session uses it. There's no need to re-pick a voice each time. Re-clone if you want a different voice; otherwise the clone stays put across all videos.
Record voice takes for a few videos to fill the recording bank
Tap Record on a Ready idea, then Voice Clone in the mode sheet
If gated: keep recording. If ready: tap Clone now (30 credits)
Wait for the percentage to hit 100% — usually under 30s
In Has Voice, tap Play take to preview, tweak sliders if needed
Tap Use this take to push the rendered audio into the caption flow