MiniMax deletes a cloned voice that isn't used within 7 days. A team that clones a voice for a seasonal campaign, a quarterly announcement, or a spokesperson who records once a month has to plan around that rule before it plans anything else.
MiniMax is a general AI company whose speech models sit beside its language, video, image, and music models, and its rapid cloning is cheap to start: $1.5 per voice, charged the first time the voice speaks. Bland Speech is a text to speech model built for phone calls, with an instant clone from about 10 seconds of audio and a professional clone for a voice that has to be exact.
Which is better, Bland Speech or MiniMax for voice cloning?
For cloned voices on phone calls, Bland Speech is the better choice as of November 2026. Bland Speech v3 ranked first of 13 AI voices in Design Arena's Audio Realism Bench results of 4 August 2026, at 1365, against 1093 for MiniMax's Speech-02 HD in 7th place, and Bland Speech costs $40 per million characters against $60 or $100 on MiniMax.
MiniMax makes more sense for a team that wants to design voices from a text description, clone across 40 languages, or hold hundreds of voices on a monthly package. A team comparing more vendors can start from the voice cloning comparison or the text to speech comparison.
Use Bland Speech when
- A team wants the voice that ranked first of 13 AI voices on Audio Realism Bench on 4 August 2026
- A brand voice should be fine-tuned on 30 minutes or more of audio, after a consent and rights review
- Clones should stay private to one organization
- A team wants one standard per-character rate, with no monthly package of audio points to manage
Use MiniMax when
- A team will design a new voice from a text description at $3 per voice
- A product will keep a cloned voice in use at least once every 7 days
- A team will manage a monthly Audio Package, from $5 for 100,000 audio points, with voice slots that rise from 10 to 800 by package
- Cloned voices should speak any of 40 languages
- Delivery will be shaped with an emotion parameter and 19 interjection tags such as laughs and sighs
What these rows mean
- Audio Realism Bench, 4 August 2026: Design Arena's blind listening test, where listeners hear two clips of the same line and pick the one that sounds more like a real person. MiniMax's entry in those results was Speech-02 HD, and every score in them is for a stock voice, not a clone. MiniMax's current speech-2.8 models weren't in those results.
- Clone types: MiniMax documents one cloning path, rapid cloning, with a minimum of 10 seconds of source audio and a maximum of 5 minutes, in mp3, m4a, or wav up to 20 MB. Bland's instant clone takes one clean sample of about 10 seconds, and its professional clone trains on 30 minutes or more.
- Charge for a cloned voice: MiniMax charges $1.5 per voice when the voice is first used for synthesis, not at cloning time. Bland charges one standard per-character rate for speech.
- Clone rules: MiniMax says a cloned voice not used within 7 days is deleted. Bland's clones are private to the customer's organization.
- Price per million characters: Each company's per-character rate for speech. MiniMax's legacy speech-2.6 and speech-02 models cost the same $100 for hd and $60 for turbo.
- First audio: Time to first audio is the wait before sound starts. MiniMax's line is its own description.
- Phone audio: Mu-law is the 8-bit format US phone networks carry.
How do Bland Speech and MiniMax clone a voice?
Bland Speech and MiniMax both clone from a short clip, and they diverge on what a clone is for. MiniMax treats a cloned voice as a cheap, disposable asset billed per voice, and Bland Speech treats it as a brand voice with a longer path behind it.
Bland Speech overview
Bland says it runs its own model and speech instead of wrapping a third party's, and Bland Speech is that speech model, sold as a studio and an API as Bland Speech, with streaming endpoints in Bland's text to speech API.
An instant clone comes from one clean sample of about 10 seconds. A professional clone trains on 30 minutes or more of audio once Bland's consent and rights review is passed. Clones are private to the customer's organization, and the customer confirms every voice is theirs to use.
Pros
- Bland Speech v3 ranked first of 13 AI voices on Audio Realism Bench on 4 August 2026, at 1365
- An instant clone from about 10 seconds, and a professional clone from 30 minutes or more
- Clones private to a customer's organization
- One standard rate, $0.04 per 1,000 characters, with no monthly package
- Mu-law and PCM output from 8 to 48 kHz for phone calls
Cons
- More than a hobby project needs, because Bland Speech is built for production calls
MiniMax overview
MiniMax calls itself "A World-Leading General AI Technology Company", and its current speech models are speech-2.8-hd and speech-2.8-turbo, released on 23 January 2026. MiniMax says that "with just a 10-second sample, Speech 2.8 precisely captures your unique texture, breathiness, and even your specific speaking pace."
Cloning is rapid: a source clip from 10 seconds to 5 minutes, and an optional prompt clip under 8 seconds that improves similarity. A cloned voice sits in a voice slot, and the number of slots depends on the Audio Package, from 10 on Starter to 800 on Business. Voice design is a separate API that builds a voice from a text description, at $3 per voice.
Pros
- Speech-02 HD ranked 7th of 13 AI voices on Audio Realism Bench on 4 August 2026
- Voice design from a text description
- 40 languages on the 2.6 and 2.8 models
- An emotion parameter and 19 interjection tags on speech-2.8 models
- G.711 mu-law at 8 kHz, and a bidirectional WebSocket that accepts text token by token and can cancel a turn
Cons
- A cloned voice is deleted if it is not used within 7 days
- $100 per million characters on speech-2.8-hd and $60 on speech-2.8-turbo
- A $1.5 charge for each cloned voice at first use
- Speech-2.8 models do not support the whisper emotion
How long does a cloned voice last?
On MiniMax, a cloned voice lasts as long as it keeps being used: unused for 7 days, it is deleted. A team that clones a campaign voice in October and needs it again in January has to call it at least weekly in between, or clone it again and pay $1.5 again. On Bland Speech, a clone is private to the organization that made it. Cloned voices are not portable between providers, so a move means re-cloning from the original recordings, with consent.
How much does a cloned voice cost to run on Bland Speech and MiniMax?
MiniMax speech costs $60 to $100 per million characters, plus $1.5 for each cloned voice, and Bland Speech costs $40 per million characters at one standard rate.
Bland Speech pricing
Bland Speech costs $0.04 per 1,000 characters, which is $40 per million, in the studio or the API, with no monthly plan. New accounts get 50,000 characters free, once, and credits never expire. A 10-minute narration costs about $0.36, and a million characters costs $40. Enterprise pricing is custom, set with Bland's sales team.
MiniMax pricing
MiniMax's API is pay as you go: speech-2.8-hd is $100 per million characters and speech-2.8-turbo is $60, for synchronous and long-form requests alike, and the legacy 2.6 and 02 models cost the same. Rapid voice cloning is $1.5 per voice and voice design is $3 per voice, each charged at first use.
MiniMax also sells monthly Audio Packages that cover every speech model, paid in audio points: Starter is $5 for 100,000 points with 10 voice slots, Standard is $30 for 300,000 with 100 slots, Pro is $99 for 1,100,000 with 250 slots, Scale is $249 for 3,300,000 with 500 slots, and Business is $999 for 20,000,000. Three-month plans are 10% off and yearly plans 20% off.
A million characters of a cloned voice costs $40 on Bland Speech, $60 on MiniMax's turbo model, and $100 on its hd model, before the $1.5 per voice.
Which sounds more human, Bland Speech or MiniMax?
Bland Speech v3 ranked first of 13 AI voices in Design Arena's Audio Realism Bench results of 4 August 2026, at 1365, and MiniMax's Speech-02 HD ranked 7th at 1093, with the human recording at 1500. MiniMax's current speech-2.8 models weren't in those results.
- #1BlandBland Speech v31365
- 2ndMicrosoftMAI-Voice-21224
- 3rdxAIGrok TTS1156
- 4thOpenAIGPT Realtime 21119
- 5thGoogleGemini 2.5 Pro TTS1109
- 7thMiniMaxSpeech-02 HD1093
Ranks are among 13 AI voices, and MiniMax's own voice is on the list. Real human recordings scored 1500, at the right edge of this scale. Bars start at 600, not zero. Source: Design Arena's published results.
A gap of 272 points is wide on a scale where listeners pick one clip over another, and MiniMax's model still placed seventh of 13, ahead of six other AI voices. The scores come from each model's most conversational stock American-English voice, so they say how human a model reads and do not say how closely a clone matches a person, as Design Arena's published results set out in the method.
How does a cloned voice handle a phone call?
Phone networks carry narrow 8 kHz audio, so a clone should be judged after that squeeze. Both return mu-law at 8 kHz: Bland Speech offers mu-law or PCM from 8 to 48 kHz and publishes 199ms to first audio, and MiniMax offers G.711 mu-law in raw or WAV form, with sample rates from 8,000 to 44,100. MiniMax's own phrase for speech-2.6-hd is "Ultra-low latency".
Why is Bland Speech the best MiniMax alternative for voice cloning?
A team that wants a cloned voice to sound human on a call, at one standard rate, without a seven-day expiry to manage, chooses Bland Speech over MiniMax: Bland Speech v3 ranked first of 13 AI voices on 4 August 2026, and clones are private to the organization. Bland is SOC 2 Type II audited and HIPAA compliant with a signed BAA, and it holds FedRAMP 20x Class C certification at Moderate, granted 17 November 2026, and Bland says speech synthesis runs inside the FedRAMP boundary.
More comparisons sit in Bland Speech vs MiniMax, the MiniMax speech alternatives list, and how to move a cloned voice to Bland Speech.
How was this comparison made?
Bland Speech and MiniMax were compared on clone types, retention, price, phone audio, and sound, using the models, rates, formats, and cloning rules each company publishes. Realism scores come from Design Arena's published results of 4 August 2026.
Frequently asked questions
Does MiniMax delete a cloned voice?
Yes, when it goes unused. MiniMax says that if a cloned voice is not used within 7 days, the system will delete it. A team can keep one alive by using it, or can clone it again.
How much does MiniMax charge to clone a voice?
MiniMax charges $1.5 per voice for rapid cloning, billed when the cloned voice is first used for synthesis, not at cloning time. Speech then costs $100 per million characters on speech-2.8-hd and $60 on speech-2.8-turbo.
How much audio does a MiniMax clone need?
MiniMax needs a source clip of at least 10 seconds and at most 5 minutes, in mp3, m4a, or wav up to 20 MB. Bland Speech makes an instant clone from one clean sample of about 10 seconds, and a professional clone from 30 minutes or more.
Is Bland Speech cheaper than MiniMax?
Per character on pay as you go, yes. Bland Speech costs $40 per million characters, and MiniMax costs $60 on speech-2.8-turbo and $100 on speech-2.8-hd, plus $1.5 for each cloned voice. MiniMax's Audio Packages bill in audio points, with Starter at $5 a month.
Does MiniMax sound as human as Bland Speech?
In Design Arena's Audio Realism Bench results of 4 August 2026, MiniMax's Speech-02 HD ranked 7th at 1093, and Bland Speech v3 ranked first of 13 AI voices at 1365. MiniMax's speech-2.8 models weren't in those results.
Can MiniMax output phone-ready audio?
Yes. MiniMax offers G.711 mu-law at 8 kHz in raw and WAV forms. Bland Speech also returns mu-law at 8 kHz, as well as PCM from 8 to 48 kHz.
