MiniMax

Published September 3, 2026 · This page indexes everything this site has verified about MiniMax and links to where each finding is set out in full. It is not a review — the review is. No listening test has been run, so nothing here says how MiniMax sounds.

On this page

What it costs

On the international platform the site records pay-as-you-go rates of $60 per 1 million characters for speech-2.8-turbo and $100 per 1 million for speech-2.8-hd, with the asynchronous long-text endpoint priced at the same rates; the mainland platform prices the same models at 2 and 3.5 CNY per 10,000 characters, and the site publishes both units as the vendor states them and performs no conversion (all checked September 1, 2026 — the review, pricing comparison). By the site’s own arithmetic from those list prices, a 100,000-character audiobook chapter comes to $6.00 on speech-2.8-turbo and $10.00 on speech-2.8-hd. No free speech tier is documented on either property, and the international audio subscription is denominated in "Audio points per month" — tiers of 100,000, 300,000, 1,100,000, 3,300,000 and 20,000,000 — a unit the documentation never defines, so no tier can be normalized to a per-character cost.

Every rate above is set against every other provider’s, normalized to one unit, in our pricing comparison.

Where MiniMax’s own material disagrees with itself

These are contradictions inside the vendor’s published documentation, not disagreements between us and the vendor. We report both readings rather than choosing one, and each links to the page where we set it out in full. We have published 13 in total; the 10 with the most bearing on a decision are below, and the rest sit in the pages linked underneath.

• MiniMax publishes an enumerated table of exactly 40 entries, identical on both platform guides, while its audio product page advertises 32 — a contradiction between two MiniMax-owned surfaces with nothing reconciling them, and no page the site read states whether the 40 are languages or locales. Where we published this.
• Setting the advertised figures against a count of the vendor’s own enumerated list, the site records that MiniMax says "40" in its documentation and "32" on its audio product page, while the enumerated list came to 40 entries — so the list matches the higher figure and the product page contradicts both. Where we published this.
• The identifiers speech-01-hd and speech-01-turbo are accepted values in the model enum of every MiniMax speech endpoint on both properties — synchronous, WebSocket, bidirectional, asynchronous and voice cloning — yet they appear in no models table and carry no published price, so the API accepts models it does not document or price. Where we published this.
• The international platform sells an audio subscription priced in "Audio points per month", but the term is never defined: the site records that the phrase occurs exactly once across the complete international documentation index it read, and never in a definition. Where we published this.
• MiniMax’s international terms impose an affirmative obligation on the customer around labelling AI-generated content, while the site’s enumeration of every property of the international text-to-audio request object found no watermark parameter at all — the mainland endpoint does expose one. The duty to label sits with the customer on the platform where the API provides no tooling to discharge it in the audio. Where we published this.
• The same company words output ownership two different ways: the international terms state that "you retain your ownership rights in Client input and generated content", while the mainland document makes ownership conditional on what applicable law provides. Where we published this.
• Earlier speech generations — speech-2.6-hd, speech-2.6-turbo, speech-02-hd and speech-02-turbo — sit inside a collapsed documentation block titled "Legacy Models" on the international property, and its Chinese equivalent on the mainland one, while remaining priced on the current pay-as-you-go page. Where we published this.
• The site found MiniMax’s published rate-limit documentation internally inconsistent in places and declined to reconcile it, reporting the published figures — 60 requests per minute for text-to-audio, 60 for voice cloning and 20 for voice design on the international property, with the mainland ceiling documented lower — as the floor to design for. Where we published this.
• The token plan’s own footnote states that coverage spans the model lineup including speech but that a small number of special models are excluded — and voice design and rapid voice cloning fall outside it, so the subscription is not a route to cheap cloning. Where we published this.
• MiniMax’s only SDK page is titled "Integrate via SDK", but its subtitle directs developers to a third party’s SDK for calling a MiniMax language model rather than a speech one; there is no MiniMax-branded speech SDK in any language, and speech integration is direct HTTP. Where we published this.

Dated and time-sensitive

Facts with a clock on them — model sunsets, policy revisions, market status. These are the entries most likely to be out of date first, which is why they carry their dates.

• 2026-03-30. MiniMax’s international Terms of Service carry an effective date of March 30, 2026. They are the document containing the ownership-and-use clause, the customer AI-disclosure duty in the user-obligations section, binding SIAC arbitration under Singapore law with no court option and no small-claims carve-out, and the statement that personal data is "stored in the data center located in the United States". Source on this site.
• 2026-09-01. No dated end-of-life exists for any MiniMax speech model on either property; the site checked for one specifically and records the absence as a contrast with several competitors. Source on this site.
• 2026-09-01. An adjacent MiniMax audio product — the music and lyrics APIs — does carry a dated service-adjustment notice, but the site records that it does not apply to speech. Source on this site.
• 2026-09-01. Cloned and designed voices are temporary by default: the API overview states verbatim that "Voices produced by cloning and voice design are temporary: the fee is charged only on first use in speech synthesis (previews within those APIs do not count)", so a clone created and never used in synthesis expires. Source on this site.
• 2026-09-02. MiniMax deletes clones that go unused — verbatim, "If a cloned voice is not used within 7 days, the system will delete it." It also documents a delete endpoint after which the identifier is burned: "Once deleted, the voice_id cannot be reused." The site records MiniMax as the only one of the ten publishing an automatic deletion rule. Source on this site.
• 2026-09-02. MiniMax sells no agent product: it documents text-to-speech, async long-form, cloning, voice design and voice management, and no realtime speech-to-speech surface — the word "Agent" appears throughout its documentation but never in a voice sense. It also attaches no latency figure to text-to-speech on either of its platforms. Source on this site.
• 2026-09-02. Reading the machine-readable documentation indexes for both the international and mainland platforms (25,188 and 26,044 bytes), the site found no match for dubbing, translation, localization, lip-sync, subtitles or captions — MiniMax publishes no dubbing product. Source on this site.
• 2026-09-01. MiniMax documents no free speech route: every published path to its audio is paid, which the site records alongside OpenAI as the only two of the ten providers with no documented free speech allowance. Source on this site.

Every page on this site about MiniMax

• Full review. MiniMax Review 2026: Two Legal Regimes, Pricing and Cloning, Docs-Verified — the complete documentation-based assessment, with sources and dates against every claim.
• Alternatives. MiniMax alternatives — organized by the reason you would leave, not as a ranked list.
• Head to head. Cartesia vs MiniMax, MiniMax vs Resemble AI.
• In our buying guides. Best AI Voice Generators in 2026: 10 Providers Compared, Best AI Dubbing Software 2026: Only Four Vendors Actually Sell It, Best Free AI Voice Generators 2026: What “Free” Actually Buys You, Most Realistic AI Voices: What Every Vendor Charges for the Claim, Best Multilingual AI Voice Generators: Every Count, Recounted, Real-Time AI Voice: Why the Latency Numbers Cannot Be Compared, Best AI Voice Cloning Software 2026: Price, Access and Consent Compared.
• In our comparisons of published data. AI Voice Generator Pricing Comparison: Every Rate Normalized, Best Text-to-Speech APIs: Limits, Transports and Auth Compared.
• In our explainers. Can You Use AI Voices Commercially? What Ten Vendors Actually Say, How AI Voice Cloning Works: What Ten Vendors Require of You, How AI Voice Generators Work, According to the Vendors Themselves.
• By what you are making. AI Voice for Audiobooks: Will Chapter 20 Match Chapter 1?, AI Voice for Business: What Survives a Procurement Review, AI Voice for eLearning: Courses Change, and the Audio Must Follow, AI Voice for Podcasts: Only Three Providers Do Two Speakers, AI Voice for YouTube: What You Must Disclose Before You Publish.

What we could not verify

Honesty about gaps beats a page that looks complete. Standing limits on everything above, as of September 3, 2026:

• How MiniMax sounds. Our listening benchmark is designed but not running. Nothing on this page or on any page it links to ranks MiniMax on audio quality, naturalness or speed.
• Anything MiniMax does not publish. Every page indexed here is built from the vendor’s own documentation. Where that documentation is silent, we record the silence rather than estimating around it.
• Whether a documented feature works as documented. We read documentation; we did not create an account or make an API call.
• Currency. Each linked page carries its own check date. This index inherits them and does not re-verify on load.

Change log

September 3, 2026 — First publication. Assembled from pages already published on this site; no new vendor claim was introduced by this page.

Published September 3, 2026 · This page shows no audio and reports no listening results, because our benchmark has not run. Corrections are recorded with their date on the corrections page. Independence note: this page contains no affiliate links, and no vendor paid for placement or influenced the order — see how we make money and our editorial policy.