streaming=true is documented as the default. false returns a generated-file URL; this is not an asynchronous discounted batch API or bidirectional input streaming.
Batch processing
Not collected
No reviewed model-specific specification; unknown is not unsupported.
Fine-tuning
Not collected
No reviewed model-specific specification; unknown is not unsupported.
Reasoning
Not collected
No reviewed model-specific specification; unknown is not unsupported.
Context caching
Not collected
No reviewed model-specific specification; unknown is not unsupported.
Base model
Not collected
No reviewed model-specific specification; unknown is not unsupported.
Select a voice by name, with optional language and accent. The public request has no neural checkpoint selector. Native engine and voice-version mapping remain unverified.Source for Model selection (opens in a new tab)
GET /auth uses Basic credentials for the registered email/password, then other calls use a Bearer access token. Access token: 3 hours. Refresh token: 30 days since last refresh; GET /auth/expire invalidates related tokens. Token lifetime is not data retention or cache lifetime.Source for Authentication (opens in a new tab)
Response format
Binary audio, or JSON with an audio URL when streaming=false. X-CereVoice-CharCount reports billed input characters. metadata=true adds X-CereVoice-Metadata linking XML word/phone/marker timings. Fixed metadata does not establish arbitrary structured-output generation.Source for Response format (opens in a new tab)
Stock voice catalog
Authenticated GET /voices filters language/accent/gender and returns name, sample rates, language/country/accent codes and regional metadata. The Heather response is an example, not the complete voice inventory or a verified neural model assignment.Source for Stock voice catalog (opens in a new tab)
Published price
The public cloud pack costs GBP 12.99 for 1,000,000 input characters; one credit equals one character. The API guide gives 10,000 free characters monthly, consumed before paid credits. Free credit resets each month. This is prepaid character billing, not tokens, audio seconds or a flat price per question. Tax, rounding, XML character counting, refunds under negotiated terms and current account availability require confirmation.Source for Published price (opens in a new tab)
Free-tier and credit-consumption order are documented separately in the API guide and /credit contract.
Creation privacy
The current API specification links Cloud Agreement V1.0, copyright 2012. A separately negotiated commercial agreement takes precedence. It reserves proprietary service rights and permits collection of technical information to improve products or customize services. It does not establish a current inference retention period, training opt-out, zero retention or processing region. Website desktop single-user restrictions are a separate licence.Source for Creation privacy (opens in a new tab)
Documentation discrepancy
The API overview prose points to POST /auth, but Swagger declares GET /auth. Desktop product deprecation is scoped separately. Linked cloud terms are V1.0 from 2012; customer commercial agreements take precedence.Source for Documentation discrepancy (opens in a new tab)
Pronunciation controls
SSML supports audio, break, emphasis, lexicon, mark, p, phoneme, prosody, say-as, sub, s and voice. meta/metadata are ignored; xml:lang and semitone prosody values are unsupported. Phoneme/lexicon tags require CereProc formats. break strength=none is unsupported.Source for Pronunciation controls (opens in a new tab)
Voice controls
Multi-voice XML, variant selection and vocal gestures. Happy/sad/calm/cross emotion tags apply only to voices with emotional support. Voice-specific behavior can change with engine updates.Source for Voice controls (opens in a new tab)
Customization options
Account-private lexicons override pronunciations for a language/accent; abbreviation files expand tokens for a language. Stored dictionaries are not base-model fine-tuning.Source for Customization options (opens in a new tab)
Access & pricing
Prices belong to an endpoint and billing tier. A provider listing does not guarantee access for every account or region.
Voice-name selection; no public model checkpoint ID.
Access
POST https://api.cerevoice.com/v2/speak, authenticated Bearer token; required plain text or XML body. streaming=true returns audio; false returns a file URL. Account and credits required.
Regions
Inference region not established; UK corporate address is not proof of data residency.
Prices retain their published currencies and billing units below.
Published billing units · original currency · observed 16 Oct 2026
Category
Usage
Rate
Unit
Prepaid character credits
Cloud credit pack1
£12.99
1,000,000 input characters
Character credit usage
Input text2
1 credit
input character
Rate conditions
One-million-credit pack; one credit is one input character. Published GBP pack price, not a token rate or per-request total. Tax, markup counting, current account offer and negotiated volume terms require confirmation.
API guide grants 10,000 free characters each month. Free credits are consumed before paid credits; free allowance resets monthly. Character metering is not a tokenizer or an audio-duration tariff.
The current API specification links Cloud Agreement V1.0, copyright 2012. A separately negotiated commercial agreement takes precedence. It reserves proprietary service rights and permits collection of technical information to improve products or customize services. It does not establish a current inference retention period, training opt-out, zero retention or processing region. Website desktop single-user restrictions are a separate licence.
Caching can reuse earlier input processing. Availability, lifetime and billing depend on the endpoint. A cache hit still generates a new response.
01 / INITIAL INPUTProcess & retain
The system processes eligible input and retains reusable state.
02 / LATER REQUESTCheck available state
Reuse depends on matching input, session state and the cache policy.
03 / CACHE HITReuse & generate
Reusing state can reduce repeated work. Any charges follow the billing rules below.
CereVoice Cloud V2
Not verified
No documented inference-cache contract. Saved audio, access tokens and stored lexicons are not prompt caching.
How it works
Not collected
No documented inference-cache contract. Saved audio, access tokens and stored lexicons are not prompt caching.
Minimum reusable input
Not collected
No documented inference-cache contract. Saved audio, access tokens and stored lexicons are not prompt caching.
Lifetime & refresh
Not collected
No documented inference-cache contract. Saved audio, access tokens and stored lexicons are not prompt caching.
How to enable it
Not collected
No documented inference-cache contract. Saved audio, access tokens and stored lexicons are not prompt caching.
Usage fields
Not collected
No documented inference-cache contract. Saved audio, access tokens and stored lexicons are not prompt caching.
Billing rules
Not collected
No documented inference-cache contract. Saved audio, access tokens and stored lexicons are not prompt caching.
Limits & data boundaries
Not collected
No documented inference-cache contract. Saved audio, access tokens and stored lexicons are not prompt caching.
No verified cache rate for this endpoint. An empty rate is not a zero charge.
Caching reviewed 16 Oct 2026
What the measurements say
No benchmark observations have been added for this model. We do not turn a vendor description into a performance score.
Latency and throughput depend on the host, request size, reasoning settings and load. There is no measured speed or global usage figure in this record.
The record behind the model
Why it is included
Proprietary cloud service with limited use rights and no published checkpoint selector. This managed component does not assert that every selected voice is CereWave or that a separate closed neural base model exists for this API operation.
