How much audio to clone a voice?seconds for a test, minutes for a good clone
Ten seconds is the usual minimum and 1 to 2 minutes of clean speech is the sweet spot for an instant clone; more than 3 minutes adds little. Professional clones that train on your voice need 30 minutes or more. On oVideo: at least 10 seconds, best 30 seconds to 2 minutes, and only the first 3 minutes are used.
Updated
Short answer
How many seconds or minutes do I need?
For an instant voice clone, 10 seconds works and 1 to 2 minutes works best. oVideo needs at least 10 seconds of speech (silence does not count), recommends 30 seconds to 2 minutes and uses the first 3 minutes. ElevenLabs, whose engine oVideo uses, gives the same advice: at least 1 minute, about 1 to 2 minutes, and more than 3 minutes yields little. Clean audio matters more than length.
- Under 10 s of speech is refused, nothing charged
- Recorder stops by itself at 3 minutes
- Upload audio or video up to 100 MB
- Voice ready in about a minute
Checked in the oVideo app on November 20, 2026.
On oVideo
Sample length on oVideo Voice Clone
- Minimum
- 10 s of speech, measured after silence at the start and end is trimmed
- Best
- 30 s to 2 min; the recorder shows “Great length” in that range
- Maximum used
- The first 3 min; recording stops by itself at 3:00
- Files
- Audio or video up to 100 MB: MP3, M4A, WAV, AAC, OGG, FLAC, MP4, MOV, MKV and more
- Time to ready
- About 1 minute, made in the background
- If it fails
- Too short, silent or blocked samples are refused and nothing is charged
- Second try
- One free re-record within 24 h
- Price
- 50 credits for a new voice; Starter $10 a month includes 1 voice slot
Checked in the oVideo app on November 20, 2026.
Compared, November 20, 2026
Sample length, tool by tool
What each tool says on its own page, read on November 20, 2026. “Not stated” means we could not find it on the pages we read. oVideo is our product, so read its row with that in mind.
| Tool | Minimum | Recommended | Upper limit | Ready in |
|---|---|---|---|---|
| oVideo Voice Clone (our product) | 10 s of speech | 30 s to 2 min | First 3 min used | About 1 min |
| ElevenLabs Instant Voice Cloning | 10 s in total (files at least 5 s) | 1 to 2 min, at least 1 min | More than 3 min adds little | Not stated on the docs page |
| ElevenLabs Professional Voice Cloning | 30 min | Closer to 2 to 3 hours | Not stated | Usually 3 to 6 hours |
| HeyGen instant clone | Not stated | About 1 min of clear speech | First 3 min used | Within seconds |
| HeyGen professional clone | 20+ min | 20+ min | Not stated | Longer, not stated |
| Synthesia | Not stated | About 20 s | Not stated | Under 15 min, by email |
| Fliki | Not stated | 30 s | Files up to 20 MB | Minutes |
| Fish Audio | 10 s | 10 s is enough, per its page | Not stated | Seconds |
| Cartesia | 10 s | Up to 60 s on Sonic 3.6 | Pro clone 30 min or more | Instant |
| MiniMax Audio | 10 s | Not stated | 5 min | Seconds, per its audio page |
| Typecast | 5 s (instant) | 20 min to 2 h (pro) | Not stated | Minutes; pro several hours |
| Canva | 10 s | Not stated | 5 min, 15 MB | A few seconds |
| Speechma | 5 s | 10 to 30 s | 30 s | Seconds |
| Filmora | Not stated | 30 s | Not stated | Seconds |
Sources: ElevenLabs Instant and Professional Voice Cloning docs, heygen.com/tool/ai-voice-cloning, synthesia.io/features/ai-voice-cloning, fliki.ai/features/voice-cloning, fish.audio/voice-clone, cartesia.ai/product/voice-cloning, platform.minimax.io voice clone guide, typecast.ai/voice-cloning, canva.com/features/ai-voice-cloning, speechma.com/voice-cloning, filmora.wondershare.com/voice-cloning.html.
Instant vs professional
Why some tools ask for seconds and others for hours
An instant clone does not train a new model. It listens to a short sample and imitates it on the fly, so 10 seconds is enough to start and a minute or two gets most of the benefit. oVideo Voice Clone, ElevenLabs Instant Voice Cloning, HeyGen’s instant clone, Fish Audio and Cartesia’s instant clone work this way.
A professional clone trains a model on your voice. ElevenLabs asks for at least 30 minutes, ideally 2 to 3 hours, and says fine tuning usually takes 3 to 6 hours; HeyGen asks for 20+ minutes; Typecast recommends 20 minutes to 2 hours. It is worth it for a full audiobook narrator voice, and more than most people need for voiceovers.
In both cases, ElevenLabs’ guide says how the audio was recorded matters more than how long it is: the AI “will attempt to mimic everything it hears”, noise included.
Get the most from a minute
Make every second count
Need words to fill the time? Our voice cloning scripts come in 30 second, 1 minute and 2 minute versions in six languages.
- One person talking. A second voice, even briefly, confuses the clone.
- No music or TV under your voice. Background noise is reduced, but it cannot all be removed.
- One steady tone for the whole sample, in the style you want back.
- Keep the same distance from the microphone, about a hand’s width.
- Uploading a long video? Only the first 3 minutes are used, so trim it to the part where you talk alone.
- Start short to test (10 to 30 seconds), then use the free re-record within 24 hours for a longer, cleaner take.
Which one should you pick?
Pick by how much audio you have
- 1Only a few seconds and you want a free test: Speechma (5 to 30 seconds, per its page).
- 2Hours of clean recordings and a full audiobook to narrate: a professional clone, such as ElevenLabs Professional Voice Cloning.
- 3A short clip of you talking, as audio or as a video from your phone: oVideo Voice Clone takes both, up to 100 MB, from Starter at $10 a month.
- 4A developer pipeline with 10 second samples: Cartesia or Fish Audio.
Checklist
How much audio, answered for oVideo
Minimum?
10 s of speech, after silence is trimmed.
Best?
30 s to 2 min.
Maximum?
Only the first 3 min are used.
Video files?
Yes, MP4, MOV and more, up to 100 MB.
How fast?
About a minute, in the background.
Too short?
Refused with “Nothing was charged”; record a little longer.
Limits and prices checked in the app on November 20, 2026.
Answers
Frequently asked questions
- How much audio is needed to clone a voice?
- For an instant clone, most tools need 10 seconds at minimum and do best with 1 to 2 minutes. On oVideo the minimum is 10 seconds of speech, 30 seconds to 2 minutes is best and only the first 3 minutes are used. Professional clones that train a model on your voice need much more: ElevenLabs asks for at least 30 minutes, ideally 2 to 3 hours.
- How many seconds of audio do you need to clone a voice?
- Ten seconds is the usual floor: oVideo, ElevenLabs, Fish Audio, Cartesia, MiniMax and Canva all state 10 seconds on their own pages (read November 20, 2026). Speechma and Typecast accept 5 seconds. Ten seconds makes a usable voice; 1 to 2 minutes makes a closer one.
- Is more audio always better?
- Not for instant clones. ElevenLabs says recording more than 3 minutes “will yield little improvement and can, in some cases, even be detrimental”, and HeyGen uses only the first three minutes. oVideo also uses only the first 3 minutes. Quality beats length: a clean, quiet minute is better than five noisy ones.
- How long does it take AI to clone a voice?
- Instant clones take seconds to a few minutes: oVideo’s is ready in about a minute, in the background, and HeyGen and Fish Audio say seconds. Synthesia says under 15 minutes by email. Professional clones take hours: ElevenLabs says fine tuning usually takes 3 to 6 hours.
- How long does it take to clone your own voice from start to finish?
- About five minutes on oVideo: a minute or two to record (or a few seconds to upload a clip), a minute for the voice to be made, then each text is spoken a moment after you press Generate voice.
- Can I clone my voice from 3 seconds of audio?
- Not on oVideo, and not on most tools: oVideo refuses less than 10 seconds of speech and nothing is charged. The shortest stated minimum we found on the tools’ own pages was 5 seconds (Speechma and Typecast instant).
- Can I clone my voice in real time?
- Not on oVideo. Cloning is done once from a recording, and speech is made from typed text, not live. Live voice changers are a different kind of tool.
- Does silence count toward the 10 seconds?
- No. oVideo trims silence at the start and end before it measures, so the 10 seconds must be speech. If you hear “We heard less than 10 s of speech”, record a little longer; nothing was charged.
- Can I upload several short clips?
- oVideo takes one recording or one file per voice. Join your clips into one file first, or simply record 30 seconds to 2 minutes in one go. ElevenLabs’ docs say the number of files does not matter there, only the total length.
- How do I tell if a voice is cloned?
- Often you can’t by ear. ElevenLabs says it embeds watermarks in the audio it generates. That is why oVideo only lets you clone your own voice, or one you have permission to use, and why YouTube asks creators to label realistic AI that makes a real person appear to say something they did not.
- Is a 10 second clone good enough?
- For a quick test, yes. For narration you will publish, record 30 seconds to 2 minutes: the clone picks up more of your rhythm and tone. On oVideo you can re-record once for free within 24 hours, so you can try short first and improve it.
- Can I use a video instead of an audio file?
- Yes. oVideo accepts MP4, MOV, M4V, MKV, WEBM and 3GP up to 100 MB, as well as audio files, and takes the voice out of the clip. Only the first 3 minutes are used, so trim to a part where you talk alone.
- How much audio does Eleven v4 need to clone a voice?
- ElevenLabs’ launch post for Eleven v4 (September 28, 2026) says instant clones can now capture a voice with high fidelity from just 10 seconds of audio. Its cloning docs still recommend 1 to 2 minutes for the best result, and its v4 page says the quality of the recording matters more than ever. oVideo Voice Clone uses Eleven v4, with a 10 second minimum and 30 seconds to 2 minutes as the best range.
Clone your voice from 30 seconds
Record or upload 30 seconds to 2 minutes, tick the consent box, and the voice is ready in about a minute. From Starter at $10 a month.
Related
- Clone my voice with AIThe full guide: steps, price, privacy.
- Voice cloning script to read30 s, 1 and 2 minute scripts in six languages.
- My voice in another languageHear yourself speak Spanish, Japanese and more.
- Best voice cloning appsThe tools compared on their own pages.
- Can ChatGPT clone my voice?What OpenAI says, and what works.
- Can CapCut clone my voice?CapCut’s clone, and when to use another tool.
- Clone your voice for YouTubeNarrate your channel in your own voice.
- Clone your voice for an audiobookLong texts, cost math and ACX rules.
- How to add AI voiceover to a videoThe full voiceover guide.
- Best AI voiceover for videoTools that voice your own video, compared.
- Eleven v4 voice cloneWhy 10 seconds can work on v4, and why clean audio matters more.
- Clone your voice on iPhoneRecord the sample in Safari.