Summary

Before you upload audio to a voice cloner, you need to know whether the sample can support a good clone. This free checker takes sample length, noise floor, peak level, room, microphone, content, processing and delivery, then returns a 0 to 100 score, a plain verdict and the three fixes worth the most points. It follows audiobook-style targets of -60 dBFS noise and -3 dBFS peaks, treats music or a second voice as a hard blocker, and runs entirely in your browser.

Is your sample ready for a voice cloner?

Score your recording before you upload it to a voice cloner. Enter length, noise floor and setup, and get a 0 to 100 score with the three fixes that matter most.

Check my sample

Check your voice cloner sample

Enter the length, noise level and recording setup of the audio you plan to upload to a voice cloner. The score and the three biggest fixes update as you type.

0 out of 100

    Working rules of thumb from voice cloning practice, not a published standard. The score estimates how much a clone has to work with, not how it will sound.

    How it works

    What the checker actually scores

    Length and level

    Sample length carries 25 points, with the top score at 3 minutes of clean speech. Peaks carry 10 more: between -12 and -3 dBFS is full marks, and a peak at 0 dBFS counts as clipped.

    Noise and space

    Noise floor carries 20 points and the room carries 15. At -60 dBFS or lower you get the full 20. A bare room or outdoor space loses most of the 15, because reverb gets cloned with the voice.

    Source and content

    Microphone, speaker content, prior processing and delivery share the last 30 points. Music or a second voice in the audio is a hard blocker: the score is capped at 49 whatever else you enter.

    Use it before every clone in four steps

    1. 1

      Record or pick the sample

      Use a dry take of one voice. Read something you would say anyway, with questions, short sentences and a few long ones.

    2. 2

      Measure two numbers

      Most DAWs and free meters show peak level and noise floor. Select a stretch of silence for the floor, and the loudest word for the peak.

    3. 3

      Read the three fixes

      The checker lists only the fixes worth the most points. Length and noise floor usually come first, so start there.

    4. 4

      Re-record, then upload

      Aim for 85 or higher. Then upload the sample to your voice cloner and generate a test line with a long sentence and a question.

    Why the sample matters more than the model

    A voice cloner copies what it hears, including what you did not mean to give it. Hiss becomes breath texture, a bare room becomes a faint reverb on every line, and a clipped peak becomes a harsh edge on loud words. Switching to a bigger model rarely fixes any of these. Fixing the sample does. The session note: record the sample at the distance you want the clone to sound, because proximity effect and room tone are copied along with the timbre.

    • About 3 minutes of clean speech gives a cloner a full pitch range.
    • Noise at -60 dBFS or lower is the audiobook ceiling and a good target for cloning.
    • Peaks between -12 and -3 dBFS leave headroom without raising the noise.
    • Music or a second voice in the sample cannot be separated cleanly.
    Hand turning the gain knob on a USB audio interface beside studio headphones

    Questions about cloning samples

    Is this a voice cloner that clones my voice here?
    No. A voice cloner needs a trained speech model running on a server, and this page runs entirely in your browser. It checks the one thing you control before any cloner touches your audio: whether the sample is good enough to clone well. Upload the sample to AnyVoice or any other cloner once the score is high.
    How long should a voice cloner sample be?
    Aim for about 3 minutes of clean speech. Instant cloners accept far less, often under a minute, but length buys stability. With 3 minutes the clone has seen your full pitch range, several sentence shapes and enough consonants to place them correctly.
    What noise floor do I need before cloning?
    Aim for -60 dBFS or lower, measured in the silence between words. That is the same ceiling audiobook distributors ask for. Anything noisier is not removed by the cloner. Hiss and room tone become part of the voice it learns.
    Why does the score penalize heavy denoising?
    Aggressive noise reduction leaves watery, warbling artifacts on consonants and breaths. A cloner treats those as part of your voice and reproduces them. A raw take with a low noise floor beats a noisy take that was cleaned afterward.
    Can I clone from a phone voice memo?
    You can, and the score will show the cost. Phone and laptop mics roll off the top end and pick up room reflections. If a phone is all you have, record 10 to 15 cm from your mouth in a closet or a room full of soft furnishings.
    Where do these weights and thresholds come from?
    They are working rules of thumb from voice cloning practice, not a published standard. The -60 dBFS noise floor and the -3 dBFS peak ceiling follow common audiobook delivery specs. The point weights reflect how much each factor tends to change clone quality, so treat the score as a prompt to fix things, not a guarantee.
    Is anything sent to a server?
    No. The scoring runs in your browser and nothing you type leaves the page, apart from the anonymous tool-run counter this site uses to see which tools get used.

    Sample ready? Clone it with AnyVoice

    AnyVoice clones a voice from a sample and gives you 8 emotion sliders to shape the delivery, so the clone is one you can steer.

    Try AnyVoice
    Clone Your Voice