Cloning a voice comes in two sizes, and the difference starts with how much of a speaker's time each one takes. An instant clone wants a short recording and is ready in minutes, while a professional clone wants a long collection of recordings, a check that the voice is the speaker's to use, and a wait for training.
Bland Speech asks for one clean sample of about 10 seconds for an instant clone. A professional clone is fine-tuned on 30 minutes or more of audio and goes through Bland's consent and rights review. ElevenLabs draws the line in the same two places, 10 seconds and about 30 minutes, as of November 2026, so the split is not a Bland quirk.
What separates an instant voice clone from a professional one?
At Bland Speech and ElevenLabs, an instant clone is built from a recording of about 10 seconds, and a professional clone is built from around 30 minutes of audio or more. Everything else follows from that gap in audio.
Mistral describes its instant method as zero-shot cloning, which means the model reproduces a voice from a short recording "without training", so the clone exists as soon as the recording does. Bland's instant clone follows the same rhythm from a customer's side: one sample, and a voice ready in about a minute.
A professional clone trains. Inworld says its professional clone "uses more audio and an asynchronous training step" and matches the original speaker more closely and more stably than an instant one. Fish Audio's flow runs upload, ownership check, training of 10 to 30 minutes, a preview, and a release, about 1 to 2 hours end to end. Cartesia says training takes up to 3 hours, and training itself is free.
Mistral goes further and calls professional cloning the recommended workflow for custom voices. Instant cloning is the quick way to hear a voice, and professional cloning is the way to build one that stays put.
How much audio does each kind need, company by company?
Bland Speech needs about 10 seconds for an instant clone and 30 minutes or more for a professional one, while Inworld starts at 3 seconds and 10 minutes, Fish Audio at 10 seconds and 10 minutes, and Typecast at 5 seconds and 5 minutes.
What these rows mean
- Instant clone: the shortest recording each company publishes for a clone made without training. ElevenLabs gives 10 seconds for Eleven v4, and Cartesia says 10 seconds is enough to start and up to 60 seconds helps retain an accent.
- Professional clone: the audio each company asks for when a voice is trained. Bland states 30 minutes or more, ElevenLabs recommends about 30 minutes, Inworld and Fish Audio set 10 minutes, and Typecast accepts a single file of 5 minutes to 3 hours.
- Plan or gate: where cloning starts on each company's plans. Bland Speech bills each character as used, so no plan sits between a business and an instant clone, and a professional clone passes a review instead.
Rime sells enterprise cloning only, ordered from Rime and trained on its Coda or Mist models, and asks for 30 to 60 minutes of clean recordings for a basic clone and two to five hours or more for the strongest result.
A business with 10 or 15 minutes of clean audio fits Inworld's or Fish Audio's professional minimum better than Bland's 30, which is a real reason to look at them.
When does a business need a professional clone instead of an instant one?
A business needs an instant clone to try a voice or run a pilot, and a professional clone when one voice has to carry a brand across a large number of calls and hold steady from the first to the last.
The reason is what a clone copies. Mistral says a clone copies more than a speaker's timbre: it picks up the emotion and pace of the recording, and also the background noise and the microphone. A sample recorded on a laptop in a kitchen brings the kitchen along. A professional set is recorded on purpose, so a business chooses the room, the microphone, and the delivery once for all of it.
Professional clones are also a limited resource at ElevenLabs and Cartesia. ElevenLabs includes 3 professional clones on Scale and 10 on Business, and Cartesia includes 2 professional slots on Startup and 4 on Scale. A team planning one voice per brand, region, or product line counts those slots before it counts the audio.
A single spokesperson voice that will run on a phone line all day is a professional job, and the extra hour of recording is small next to a year of calls in a voice that wobbles.
What does consent look like for each kind?
At Bland Speech, a customer confirms that each voice is theirs to use, and a professional clone also goes through Bland's consent and rights review, while ElevenLabs runs a voice captcha and Fish Audio verifies ownership before training.
ElevenLabs' professional clone verification asks the voice owner to read lines from a captcha image aloud and record them, and ElevenLabs says the step exists to confirm permission to use the voice. Fish Audio asks for the same kind of proof before it starts training, and a clone that fails verification is stopped there. Bland's review sits at the same point in the process: before a professional voice is trained.
The check grows with the kind of clone because the stakes grow with it. A professional voice is the one a business will build a phone line around. Consent rules by provider, with the evidence for each, are in Is voice cloning legal for business?, and the best voice cloning with consent verification ranks providers on that check alone.
Which kind of voice cloning does Bland Speech fit?
Bland Speech fits a business that wants to hear a cloned voice within a minute and move to a professional clone when that voice has to carry real calls, with every clone private to the customer's organization.
Instant cloning on Bland Speech runs on BTTS V3 by default, which Bland describes as its highest-fidelity engine, from a single sample of about 10 seconds. Speech is billed at $0.04 per 1,000 characters, pay as you go, with 50,000 free characters once for a new account. Where a team wants the voice to be trained, a professional clone takes 30 minutes or more of audio and the review above. A voice cloned somewhere else does not move across, and the guide on moving a cloned voice to Bland Speech covers re-cloning from the original recordings with the speaker's consent.
Bland Speech is built for phone calls, and the text to speech API delivers audio at 8 to 48 kHz and in mu-law, a compressed encoding used on phone lines. The voice cloning comparison hub and the wider text to speech comparison set Bland Speech against each provider above, and Bland Speech has a studio for trying a voice before building on it.
Frequently asked questions
How long does an instant voice clone take?
A Bland Speech instant clone is ready in about a minute from one sample of around 10 seconds, according to Bland. Professional clones take longer at every company: Fish Audio's flow runs 1 to 2 hours with 10 to 30 minutes of training, Inworld says professional training typically takes 30 to 60 minutes, and Cartesia says training takes up to 3 hours.
Can a 10-second clone sound like the speaker on every call?
A 10-second clone can capture the voice, but it also copies the delivery and recording conditions of the sample, so results follow the recording. Cartesia says 10 seconds is enough to get started and up to 60 seconds better retains the speaker's accent, and Inworld says the longer the sample, up to its 30-second limit, the better the speaker similarity.
Does a professional clone need a recording studio?
Providers describe the recordings as clean and single-speaker. Mistral recommends 30 minutes to 3 hours of clean, high-quality audio featuring only one speaker, Fish Audio asks for at least 10 minutes of clean audio of one speaker, and Murf asks for studio-grade recordings of up to 90 minutes for its Enterprise professional clones.
Does an instant clone need consent too?
Yes, at the providers that publish their steps. Bland Speech has a customer confirm that every voice is theirs to use, Google's Instant Custom Voice required a recorded consent statement, and Speechify requires a spoken challenge phrase that it checks and keeps as the consent record.
Are cloned voices shared with other customers?
Not at Bland Speech: clones are private to the customer's organization. Fish Audio works differently for released professional voices, which can earn a share when other creators pay to use them, so a business reads that setting before releasing a voice.
Can a cloned voice move from one provider to another?
A cloned voice is not portable, so a business re-clones from the original recordings, with the speaker's consent. Typecast, for example, bars exporting a cloned voice outside its APIs.
