Managed speech API spanning voice-quality tiers, with streaming, complete-file generation and word/viseme events. The public voice list is Classic-only.
Mode-specific character limits, not tokens. No overall POST file/events maximum established; do not apply the stream limit to every mode.
TTS text limits
GET text is limited to 2,048 characters. Streaming is limited to 3,000, so GET streaming must satisfy both. Public demo configuration separately sets 500 characters and 15 commands/day, with max_play_command_per_day=30. These UI quotas do not establish production API throughput or concurrency.Source for TTS text limits (opens in a new tab)
GET restriction comes from the API contract; the configuration explicitly confirms the streaming and demo limits.
Opus requires 48,000 Hz. Supported bitrate/rate combinations depend on codec. API output rates do not prove a native model training sample rate.
Voice controls
API speed 30–300 (default 100), shaping 50–150 (100), volume 50–65535 (32768). Text tags have separate ranges/defaults, including volume 0–65535 (65535). Changing voices resets pitch/speed/volume defaults.Source for Voice controls (opens in a new tab)
The tags page documents different volume semantics. Do not silently combine API controls with inline tags.
Response format
output=stream returns audio chunks; file returns complete audio or audio+event ZIP; events returns event JSON. wordpos/mouthpos/markpos enable word, viseme and marker events. Streamed events/audio use type:size followed by a newline and bytes; a logical buffer can span transport chunks. WordOffset/WordSize are bytes, not Unicode character indexes.Source for Response format (opens in a new tab)
Authentication
POST /api/login/ with email/password returns a token; use Authorization: Token <token>. Documentation says tokens have no expiration until logout invalidates them. A token lifetime is not a cache or data-retention policy.Source for Authentication (opens in a new tab)
Pronunciation controls
Comma-separated .dic dictionaries; when entries collide, the last loaded dictionary wins. Inline phonetic tags use a language-dependent alphabet and are documented for single words. Separate SSML guide is linked, but full SSML compatibility has not been verified here.Source for Pronunciation controls (opens in a new tab)
Source audio handling
Audio insertion tags require raw 16-bit mono PCM at the voice sample rate; uploaded MP3/WAV can be converted. This insertion restriction does not limit all generated output to raw PCM. Uploaded sounds and dictionary files are account storage, not model fine-tuning.Source for Source audio handling (opens in a new tab)
Published price
Provider offers prepaid, postpaid and subscription arrangements and describes audio-duration credit billing. No current numeric API tariff by voice quality/transport is established. UI strings differ on whether every preview or only generated/downloaded audio is charged; confirm account contract. No token or per-question conversion.Source for Published price (opens in a new tab)
Model terms
Cloud terms last updated 09-09-26 distinguish one-use streaming audio, which cannot be recorded/reused/redistributed, from licensed audio-production files. The published default licence is personal/noncommercial; commercials, public/transport announcements and broadcast uses require a separate agreement. Customer-specific permissions govern the actual application. No commercial licence is inferred from paid access.Source for Model terms (opens in a new tab)
Creation privacy
The linked Acapela privacy policy permits processing voice data for product/AI development and improvement where authorized, and necessary third-party processing. It gives a general five-year inactivity deletion statement but separate AI-recording/documentation retention provisions. Exact inference retention, training opt-out and processing region are not established; on-device My-Own-Voice claims do not apply automatically to cloud generation.Source for Creation privacy (opens in a new tab)
Documentation discrepancy
Public /api/voice/ currently returns only 124 standard-classic _HQ entries, while product pages advertise Neural/Premium/Standard. Authenticated /api/account/ is authoritative for entitlements. Some samples use output=storage although the parameter table specifies file. Changelog dates for 14.000.3 and 12.000.9 are out of sequence; do not invent corrected dates.Source for Documentation discrepancy (opens in a new tab)
124 public standard-classic voice IDs; all names end in _HQ and version fields are null. This is a captured public subset, not all account-specific Neural/Premium/custom voices.Source for Stock voice catalog (opens in a new tab)
Voice-name selection; no standalone model checkpoint selector.
Access
GET or POST https://www.acapela-cloud.com/api/command/ with voice, text, output=stream and Authorization: Token. POST stream limit 3,000 characters; GET limit 2,048. Mixed audio/events require framing-aware parsing.
Regions
Processing region unverified; Belgian company address is not inference residency.
Endpoint pricing has not been collected. Check the provider for applicable rates.
Data & enterprise terms
Cloud terms last updated 09-09-26 distinguish one-use streaming audio, which cannot be recorded/reused/redistributed, from licensed audio-production files. The published default licence is personal/noncommercial; commercials, public/transport announcements and broadcast uses require a separate agreement. Customer-specific permissions govern the actual application. No commercial licence is inferred from paid access. The linked Acapela privacy policy permits processing voice data for product/AI development and improvement where authorized, and necessary third-party processing. It gives a general five-year inactivity deletion statement but separate AI-recording/documentation retention provisions. Exact inference retention, training opt-out and processing region are not established; on-device My-Own-Voice claims do not apply automatically to cloud generation.
Voice-name selection; no standalone model checkpoint selector.
Access
GET or POST https://www.acapela-cloud.com/api/command/ with voice, text, output=file and Authorization: Token. Complete audio or ZIP containing audio and enabled events. POST file text limit not established; GET remains 2,048 characters.
Regions
Processing region unverified; Belgian company address is not inference residency.
Endpoint pricing has not been collected. Check the provider for applicable rates.
Data & enterprise terms
Cloud terms last updated 09-09-26 distinguish one-use streaming audio, which cannot be recorded/reused/redistributed, from licensed audio-production files. The published default licence is personal/noncommercial; commercials, public/transport announcements and broadcast uses require a separate agreement. Customer-specific permissions govern the actual application. No commercial licence is inferred from paid access. The linked Acapela privacy policy permits processing voice data for product/AI development and improvement where authorized, and necessary third-party processing. It gives a general five-year inactivity deletion statement but separate AI-recording/documentation retention provisions. Exact inference retention, training opt-out and processing region are not established; on-device My-Own-Voice claims do not apply automatically to cloud generation.
Voice-name selection; no standalone model checkpoint selector.
Access
GET or POST https://www.acapela-cloud.com/api/command/ with voice, text, output=events and Authorization: Token. Timing/viseme event JSON; no audio requested in this mode. POST event text limit not established; GET remains 2,048 characters.
Regions
Processing region unverified; Belgian company address is not inference residency.
Endpoint pricing has not been collected. Check the provider for applicable rates.
Data & enterprise terms
Cloud terms last updated 09-09-26 distinguish one-use streaming audio, which cannot be recorded/reused/redistributed, from licensed audio-production files. The published default licence is personal/noncommercial; commercials, public/transport announcements and broadcast uses require a separate agreement. Customer-specific permissions govern the actual application. No commercial licence is inferred from paid access. The linked Acapela privacy policy permits processing voice data for product/AI development and improvement where authorized, and necessary third-party processing. It gives a general five-year inactivity deletion statement but separate AI-recording/documentation retention provisions. Exact inference retention, training opt-out and processing region are not established; on-device My-Own-Voice claims do not apply automatically to cloud generation.
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.
Acapela Cloud
Not verified
No documented inference-cache contract. Downloaded voices, stored dictionaries, reusable production audio and browser caching are not prompt caching.
How it works
Not collected
No documented inference-cache contract. Downloaded voices, stored dictionaries, reusable production audio and browser caching are not prompt caching.
Minimum reusable input
Not collected
No documented inference-cache contract. Downloaded voices, stored dictionaries, reusable production audio and browser caching are not prompt caching.
Lifetime & refresh
Not collected
No documented inference-cache contract. Downloaded voices, stored dictionaries, reusable production audio and browser caching are not prompt caching.
How to enable it
Not collected
No documented inference-cache contract. Downloaded voices, stored dictionaries, reusable production audio and browser caching are not prompt caching.
Usage fields
Not collected
No documented inference-cache contract. Downloaded voices, stored dictionaries, reusable production audio and browser caching are not prompt caching.
Billing rules
Not collected
No documented inference-cache contract. Downloaded voices, stored dictionaries, reusable production audio and browser caching are not prompt caching.
Limits & data boundaries
Not collected
No documented inference-cache contract. Downloaded voices, stored dictionaries, reusable production audio and browser caching are not prompt caching.
No verified cache rate for this endpoint. An empty rate is not a zero charge.
Caching reviewed 16 Oct 2026
Acapela Cloud
Not verified
No documented inference-cache contract. Downloaded voices, stored dictionaries, reusable production audio and browser caching are not prompt caching.
How it works
Not collected
No documented inference-cache contract. Downloaded voices, stored dictionaries, reusable production audio and browser caching are not prompt caching.
Minimum reusable input
Not collected
No documented inference-cache contract. Downloaded voices, stored dictionaries, reusable production audio and browser caching are not prompt caching.
Lifetime & refresh
Not collected
No documented inference-cache contract. Downloaded voices, stored dictionaries, reusable production audio and browser caching are not prompt caching.
How to enable it
Not collected
No documented inference-cache contract. Downloaded voices, stored dictionaries, reusable production audio and browser caching are not prompt caching.
Usage fields
Not collected
No documented inference-cache contract. Downloaded voices, stored dictionaries, reusable production audio and browser caching are not prompt caching.
Billing rules
Not collected
No documented inference-cache contract. Downloaded voices, stored dictionaries, reusable production audio and browser caching are not prompt caching.
Limits & data boundaries
Not collected
No documented inference-cache contract. Downloaded voices, stored dictionaries, reusable production audio and browser caching are not prompt caching.
No verified cache rate for this endpoint. An empty rate is not a zero charge.
Caching reviewed 16 Oct 2026
Acapela Cloud
Not verified
No documented inference-cache contract. Downloaded voices, stored dictionaries, reusable production audio and browser caching are not prompt caching.
How it works
Not collected
No documented inference-cache contract. Downloaded voices, stored dictionaries, reusable production audio and browser caching are not prompt caching.
Minimum reusable input
Not collected
No documented inference-cache contract. Downloaded voices, stored dictionaries, reusable production audio and browser caching are not prompt caching.
Lifetime & refresh
Not collected
No documented inference-cache contract. Downloaded voices, stored dictionaries, reusable production audio and browser caching are not prompt caching.
How to enable it
Not collected
No documented inference-cache contract. Downloaded voices, stored dictionaries, reusable production audio and browser caching are not prompt caching.
Usage fields
Not collected
No documented inference-cache contract. Downloaded voices, stored dictionaries, reusable production audio and browser caching are not prompt caching.
Billing rules
Not collected
No documented inference-cache contract. Downloaded voices, stored dictionaries, reusable production audio and browser caching are not prompt caching.
Limits & data boundaries
Not collected
No documented inference-cache contract. Downloaded voices, stored dictionaries, reusable production audio and browser caching are not prompt caching.
No verified cache rate for this endpoint. An empty rate is not a zero charge.
Caching reviewed 16 Oct 2026
Voice choices
124 voices
Choose a voice within Acapela Cloud Speech. These are API voice identifiers, not additional model releases. Source links document each voice.
Published voice catalog
Voice / API ID
Language & accent
Delivery
LeilaLeila22k_HQ
Arabic ara-sau
Classic; Adult; mood Neutral; provider type standard-classic. Version not provided. Public metadata is not proof of account entitlement or neural compatibility.Voice source (opens in a new tab)
MehdiMehdi22k_HQ
Arabic ara-sau
Classic; Adult; mood Neutral; provider type standard-classic. Version not provided. Public metadata is not proof of account entitlement or neural compatibility.Voice source (opens in a new tab)
NizarNizar22k_HQ
Arabic ara-sau
Classic; Adult; mood Neutral; provider type standard-classic. Version not provided. Public metadata is not proof of account entitlement or neural compatibility.Voice source (opens in a new tab)
SalmaSalma22k_HQ
Arabic ara-sau
Classic; Adult; mood Neutral; provider type standard-classic. Version not provided. Public metadata is not proof of account entitlement or neural compatibility.Voice source (opens in a new tab)
NizarengNizareng22k_HQ
ArabicEnglish eng-ara
Classic; Adult; mood Neutral; provider type standard-classic. Version not provided. Public metadata is not proof of account entitlement or neural compatibility.Voice source (opens in a new tab)
LisaLisa22k_HQ
AustralianEnglish eng-aus
Classic; Adult; mood Neutral; provider type standard-classic. Version not provided. Public metadata is not proof of account entitlement or neural compatibility.Voice source (opens in a new tab)
TylerTyler22k_HQ
AustralianEnglish eng-aus
Classic; Adult; mood Neutral; provider type standard-classic. Version not provided. Public metadata is not proof of account entitlement or neural compatibility.Voice source (opens in a new tab)
JeroenJeroen22k_HQ
BelgianDutch nld-bel
Classic; Adult; mood Neutral; provider type standard-classic. Version not provided. Public metadata is not proof of account entitlement or neural compatibility.Voice source (opens in a new tab)
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 hosted service with voice-quality tiers and account entitlements. This managed record does not assert that every Classic voice is a closed neural checkpoint or duplicate the Neural TTS family as another base model.
