Proprietary CPU speech-synthesis family for licensed server, desktop and embedded deployment. Regional packages have different hardware and voice specifications.
16-bit linear PCM, 16-bit linear PCM WAV, 8-bit A-law PCM, 8-bit A-law PCM WAV, 8-bit mu-law PCM, 8-bit mu-law PCM WAV, 8-bit unsigned linear PCM WAV, 4-bit Dialogic ADPCMSource for Output formats (opens in a new tab)
Embedded SDK formats; do not inherit speechCloud codecs/sample rates into local deployments.
Voice controls
Speaking rate, pitch and volume; SSML prosody, pauses, phonetic input and inline voice/language switching. Multiple pronunciation dictionaries per language and IPA transcription are supported.Source for Voice controls (opens in a new tab)
SDK support
Embedded Linux: C/C++; iOS: Objective-C; Android: Java. The specification also names WinCE under bindings but omits it from its current OS list, so current WinCE support requires confirmation.Source for SDK support (opens in a new tab)
Model terms
Separate licence defines platform, application/device, voices, duration and permitted context. A licence file enforces voices and term; synthesis-speed restrictions can also apply. Licensed on-device delivery is not an open-weight release.Source for Model terms (opens in a new tab)
Published price
Not collected
No numeric SDK tariff collected. Prices and rights depend on the platform, application, licensed voices and term.
Creation privacy
Provider Privacy Notice v1.4 (April 3, 2026) states voice-model improvement uses actor-recorded data, not customer input. It also discloses third-party engines depending on selected voice. Its controller notice excludes embedded-service processing where ReadSpeaker acts as processor; the integrating organization and applicable DPA govern that use. Exact inference retention, opt-out controls and processing location remain unverified. EU corporate address is not a residency guarantee.Source for Creation privacy (opens in a new tab)
168 regional display profiles. Names are preserved from the provider table; these are not verified API selectors or 168 independent base models. Global SDK documents separately state 150+ voices in 50+ languages.
Global 2026 SDK specifications and Korean product tables give different hardware/memory figures. Treat them as distinct delivery contracts, not universal limits or a single corrected value. SDK 4.5 and regional ReadSpeaker v3.0 software certification do not establish model checkpoint revisions.Source for Documentation discrepancy (opens in a new tab)
Global SDK documentation is separately linked. No hidden version mapping inferred.
Model selection
Not collected
SDK voice files and regional profiles are licensed delivery items. Marketing examples are explicitly illustrative, not authoritative API identifiers.
Access & pricing
Prices belong to an endpoint and billing tier. A provider listing does not guarantee access for every account or region.
Licensed voice files; exact delivered SDK selector/version must be confirmed.
Access
Licensed on-device SDK delivered for the ordered platform. Exact API signatures and voice names come with the customer package; no hosted public inference route.
Regions
Local inference on the licensed deployment; enrollment, support and delivery can use provider systems.
Endpoint pricing has not been collected. Check the provider for applicable rates.
Data & enterprise terms
Licence determines platform, context, term and voices. Provider Privacy Notice v1.4 (April 3, 2026) states voice-model improvement uses actor-recorded data, not customer input. It also discloses third-party engines depending on selected voice. Its controller notice excludes embedded-service processing where ReadSpeaker acts as processor; the integrating organization and applicable DPA govern that use. Exact inference retention, opt-out controls and processing location remain unverified. EU corporate address is not a residency guarantee.
Licensed voice files; exact delivered SDK selector/version must be confirmed.
Access
Licensed on-device SDK delivered for the ordered platform. Exact API signatures and voice names come with the customer package; no hosted public inference route.
Regions
Local inference on the licensed deployment; enrollment, support and delivery can use provider systems.
Endpoint pricing has not been collected. Check the provider for applicable rates.
Data & enterprise terms
Licence determines platform, context, term and voices. Provider Privacy Notice v1.4 (April 3, 2026) states voice-model improvement uses actor-recorded data, not customer input. It also discloses third-party engines depending on selected voice. Its controller notice excludes embedded-service processing where ReadSpeaker acts as processor; the integrating organization and applicable DPA govern that use. Exact inference retention, opt-out controls and processing location remain unverified. EU corporate address is not a residency guarantee.
Licensed voice files; exact delivered SDK selector/version must be confirmed.
Access
Korean product: Windows/Linux/Embedded Linux; ARM 64-bit at 2 GHz+ or x64 at 1.6 GHz+, with AVX requirement as stated. RAM 12 MB, engine 7–32 MB. Package-specific integration and licence required; no public inference URL established.
Regions
Local inference on the licensed deployment; enrollment, support and delivery can use provider systems.
Endpoint pricing has not been collected. Check the provider for applicable rates.
Data & enterprise terms
Licence determines platform, context, term and voices. Provider Privacy Notice v1.4 (April 3, 2026) states voice-model improvement uses actor-recorded data, not customer input. It also discloses third-party engines depending on selected voice. Its controller notice excludes embedded-service processing where ReadSpeaker acts as processor; the integrating organization and applicable DPA govern that use. Exact inference retention, opt-out controls and processing location remain unverified. EU corporate address is not a residency guarantee.
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.
ReadSpeaker speechEngine SDK Embedded
Not verified
No reviewed inference-cache contract. Downloaded voice files, dictionaries, generated-audio reuse and runtime memory are not prompt caching.
How it works
Not collected
No reviewed inference-cache contract. Downloaded voice files, dictionaries, generated-audio reuse and runtime memory are not prompt caching.
Minimum reusable input
Not collected
No reviewed inference-cache contract. Downloaded voice files, dictionaries, generated-audio reuse and runtime memory are not prompt caching.
Lifetime & refresh
Not collected
No reviewed inference-cache contract. Downloaded voice files, dictionaries, generated-audio reuse and runtime memory are not prompt caching.
How to enable it
Not collected
No reviewed inference-cache contract. Downloaded voice files, dictionaries, generated-audio reuse and runtime memory are not prompt caching.
Usage fields
Not collected
No reviewed inference-cache contract. Downloaded voice files, dictionaries, generated-audio reuse and runtime memory are not prompt caching.
Billing rules
Not collected
No reviewed inference-cache contract. Downloaded voice files, dictionaries, generated-audio reuse and runtime memory are not prompt caching.
Limits & data boundaries
Not collected
No reviewed inference-cache contract. Downloaded voice files, dictionaries, generated-audio reuse and runtime memory are not prompt caching.
No verified cache rate for this endpoint. An empty rate is not a zero charge.
Caching reviewed 16 Oct 2026
ReadSpeaker speechEngine SDK
Not verified
No reviewed inference-cache contract. Downloaded voice files, dictionaries, generated-audio reuse and runtime memory are not prompt caching.
How it works
Not collected
No reviewed inference-cache contract. Downloaded voice files, dictionaries, generated-audio reuse and runtime memory are not prompt caching.
Minimum reusable input
Not collected
No reviewed inference-cache contract. Downloaded voice files, dictionaries, generated-audio reuse and runtime memory are not prompt caching.
Lifetime & refresh
Not collected
No reviewed inference-cache contract. Downloaded voice files, dictionaries, generated-audio reuse and runtime memory are not prompt caching.
How to enable it
Not collected
No reviewed inference-cache contract. Downloaded voice files, dictionaries, generated-audio reuse and runtime memory are not prompt caching.
Usage fields
Not collected
No reviewed inference-cache contract. Downloaded voice files, dictionaries, generated-audio reuse and runtime memory are not prompt caching.
Billing rules
Not collected
No reviewed inference-cache contract. Downloaded voice files, dictionaries, generated-audio reuse and runtime memory are not prompt caching.
Limits & data boundaries
Not collected
No reviewed inference-cache contract. Downloaded voice files, dictionaries, generated-audio reuse and runtime memory are not prompt caching.
No verified cache rate for this endpoint. An empty rate is not a zero charge.
Caching reviewed 16 Oct 2026
ReadSpeaker Korea licensed package
Not verified
No reviewed inference-cache contract. Downloaded voice files, dictionaries, generated-audio reuse and runtime memory are not prompt caching.
How it works
Not collected
No reviewed inference-cache contract. Downloaded voice files, dictionaries, generated-audio reuse and runtime memory are not prompt caching.
Minimum reusable input
Not collected
No reviewed inference-cache contract. Downloaded voice files, dictionaries, generated-audio reuse and runtime memory are not prompt caching.
Lifetime & refresh
Not collected
No reviewed inference-cache contract. Downloaded voice files, dictionaries, generated-audio reuse and runtime memory are not prompt caching.
How to enable it
Not collected
No reviewed inference-cache contract. Downloaded voice files, dictionaries, generated-audio reuse and runtime memory are not prompt caching.
Usage fields
Not collected
No reviewed inference-cache contract. Downloaded voice files, dictionaries, generated-audio reuse and runtime memory are not prompt caching.
Billing rules
Not collected
No reviewed inference-cache contract. Downloaded voice files, dictionaries, generated-audio reuse and runtime memory are not prompt caching.
Limits & data boundaries
Not collected
No reviewed inference-cache contract. Downloaded voice files, dictionaries, generated-audio reuse and runtime memory 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
ReadSpeaker explicitly identifies proprietary neural technology; regional product pages and licensed SDK specifications identify these named neural families. These are family records, not an assertion of public weight downloads or immutable checkpoints. Third-party engines disclosed for some hosted voices are not reattributed to ReadSpeaker.
