Skip to content
Model infoModelsSnapshot: 15 Oct 2026

NAVER Cloud / CLOVA Speech

CLOVA Speech (NEST)

NEST speech transcription with short-file, long-file and live gRPC modes. Includes long-file batch jobs, speaker labels and word alignment where documented.

availableClosed weightsSpeech recognition
Official documentation (opens in a new tab)Reviewed 13 Oct 2026
12 of 23 specification fields collected
Audio durationShort files: 60 seconds. Long files: 2 hours sync or 6 hours async/batch.See specification notes below
File size limitShort files: 10 MB. Long files: 2 GB.See specification notes below
Input billingNot collectedCLOVA Speech
Output billingNot collectedNo inferred price

Specifications

Provider-documented facts. “Not collected” means we have not verified that field yet, not that the feature is unsupported.

Context / input limit
Not collected
Maximum output
Not collected
Input modalities
Audio, VideoSource for Input modalities (opens in a new tab)

Video containers are supported by long-file recognition. Only the audio track is transcribed; this is not video understanding.

Reasoning
Not collected
Tool calling
Not collected
Structured output
Not collected
Streaming
SupportedSource for Streaming (opens in a new tab)

Live recognition uses a separate gRPC endpoint. File transcription is not the same transport.

Batch processing
SupportedSource for Batch processing (opens in a new tab)

Explicit multi-file workflow using Object Storage; one batch per Basic domain and at most ten per account. Free domains cannot create batches. No discount is inferred.

Fine-tuning
Not collected
Knowledge cutoff
Not collected
Release date
Not collected
Parameters
Not collected
Architecture
Not collected
Languages
Korean, English, Japanese, Chinese (Simplified), Chinese (Traditional), Korean/English simultaneous recognitionSource for Languages (opens in a new tab)

This is a union across modes. Live gRPC supports only Korean, English and Japanese. Traditional Chinese and mixed Korean/English are documented for long-file recognition.

Context caching
Not collected
Audio duration limits
Short files: 60 seconds. Long files: 2 hours sync or 6 hours async/batch.Source for Audio duration limits (opens in a new tab)

The live gRPC session lifetime is not established: both the English and Korean API error tables leave its numeric threshold blank.

Maximum audio file
Short files: 10 MB. Long files: 2 GB.Source for Maximum audio file (opens in a new tab)
Speaker diarization
SupportedSource for Speaker diarization (opens in a new tab)

Long-file recognition: diarization.enable defaults to true. This is not a statement about the short-file or gRPC modes.

Word timestamps
SupportedSource for Word timestamps (opens in a new tab)

Long-file wordAlignment is enabled by default.

Output formats
JSON, SRT, SMISource for Output formats (opens in a new tab)

Long-file results. These fixed response formats do not establish caller-defined JSON-schema generation.

Keyterm prompting limits
Long files: up to 1,000 Korean/English boosting entries.Source for Keyterm prompting limits (opens in a new tab)

Short-file limits differ and are recorded on that endpoint. Keyword boosting is not evidence of fine-tuning.

Usage metering
Billed by recognition duration. Numeric rates and any feature surcharges remain unverified.Source for Usage metering (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.

API model ID
Not collected

This service exposes a dedicated route or workflow rather than a selectable checkpoint ID.

Access
CLOVA Speech domain and secret key required. Domain limits and permissions apply.
Regions
Korea; Classic and VPC environments are documented.
API endpoint URL
POST {domain invoke URL}/recognizer/object-storageSource for API endpoint URL (opens in a new tab)
Languages
ko-KR, en-US, enko, ja, zh-cn, zh-twSource for Languages (opens in a new tab)
Audio duration limits
2 hours synchronous; 6 hours asynchronous.Source for Audio duration limits (opens in a new tab)
Input audio format
View all 13 entries
  • MP3
  • AAC
  • AC3
  • OGG
  • FLAC
  • WAV
  • M4A
  • AVI
  • MP4
  • MOV
  • WMV
  • FLV
  • MKV
Source for Input audio format (opens in a new tab)
Request mode
sync returns recognition results; async requires callback URL or resultToObs. Async is the default.Source for Request mode (opens in a new tab)
Speaker diarization
SupportedSource for Speaker diarization (opens in a new tab)

Enabled by default; may be disabled.

Keyterm prompting limits
1,000 entries, Korean and English only. boostings and useDomainBoostings are mutually exclusive.Source for Keyterm prompting limits (opens in a new tab)

Endpoint pricing has not been collected. Check the provider for applicable rates.

Data & enterprise terms

Retention, training use, regional processing and enterprise terms have not been summarized for this endpoint.

API model ID
Not collected

This service exposes a dedicated route or workflow rather than a selectable checkpoint ID.

Access
CLOVA Speech domain and secret key required. Domain limits and permissions apply.
Regions
Korea; Classic and VPC environments are documented.
API endpoint URL
POST {domain invoke URL}/recognizer/urlSource for API endpoint URL (opens in a new tab)
Languages
ko-KR, en-US, enko, ja, zh-cn, zh-twSource for Languages (opens in a new tab)
Audio duration limits
2 hours synchronous; 6 hours asynchronous.Source for Audio duration limits (opens in a new tab)
Input audio format
View all 13 entries
  • MP3
  • AAC
  • AC3
  • OGG
  • FLAC
  • WAV
  • M4A
  • AVI
  • MP4
  • MOV
  • WMV
  • FLV
  • MKV
Source for Input audio format (opens in a new tab)
Request mode
sync returns recognition results; async requires callback URL or resultToObs. Async is the default.Source for Request mode (opens in a new tab)
Speaker diarization
SupportedSource for Speaker diarization (opens in a new tab)

Enabled by default; may be disabled.

Keyterm prompting limits
1,000 entries, Korean and English only. boostings and useDomainBoostings are mutually exclusive.Source for Keyterm prompting limits (opens in a new tab)

Endpoint pricing has not been collected. Check the provider for applicable rates.

Data & enterprise terms

Retention, training use, regional processing and enterprise terms have not been summarized for this endpoint.

API model ID
Not collected

This service exposes a dedicated route or workflow rather than a selectable checkpoint ID.

Access
CLOVA Speech domain and secret key required. Domain limits and permissions apply.
Regions
Korea; Classic and VPC environments are documented.
API endpoint URL
POST {domain invoke URL}/recognizer/uploadSource for API endpoint URL (opens in a new tab)
Languages
ko-KR, en-US, enko, ja, zh-cn, zh-twSource for Languages (opens in a new tab)
Audio duration limits
2 hours synchronous; 6 hours asynchronous.Source for Audio duration limits (opens in a new tab)
Input audio format
View all 13 entries
  • MP3
  • AAC
  • AC3
  • OGG
  • FLAC
  • WAV
  • M4A
  • AVI
  • MP4
  • MOV
  • WMV
  • FLV
  • MKV
Source for Input audio format (opens in a new tab)
Request mode
sync returns recognition results; async requires callback URL or resultToObs. Async is the default.Source for Request mode (opens in a new tab)
Speaker diarization
SupportedSource for Speaker diarization (opens in a new tab)

Enabled by default; may be disabled.

Keyterm prompting limits
1,000 entries, Korean and English only. boostings and useDomainBoostings are mutually exclusive.Source for Keyterm prompting limits (opens in a new tab)

Endpoint pricing has not been collected. Check the provider for applicable rates.

Data & enterprise terms

Retention, training use, regional processing and enterprise terms have not been summarized for this endpoint.

API model ID
Not collected

This service exposes a dedicated route or workflow rather than a selectable checkpoint ID.

Access
CLOVA Speech domain and secret key required. Domain limits and permissions apply.
Regions
Korea; Classic and VPC environments are documented.
API endpoint URL
POST https://clovaspeech-gw.ncloud.com/recog/v1/sttSource for API endpoint URL (opens in a new tab)
Audio formats
MP3, AAC, AC3, OGG, FLAC, WAVSource for Audio formats (opens in a new tab)

The broader prerequisites page also lists M4A, but this API reference does not.

Pronunciation assessment
Optional pronunciation assessment for Korean/English, reference utterance and waveform graph. Korean keyword boosting: up to 512 characters total; keywords need at least three characters.Source for Pronunciation assessment (opens in a new tab)

Endpoint pricing has not been collected. Check the provider for applicable rates.

Data & enterprise terms

Retention, training use, regional processing and enterprise terms have not been summarized for this endpoint.

API model ID
Not collected

This service exposes a dedicated route or workflow rather than a selectable checkpoint ID.

Access
CLOVA Speech domain and secret key required. Domain limits and permissions apply.
Regions
Korea; Classic and VPC environments are documented.
API endpoint URL
gRPC over TLS: clovaspeech-gw.ncloud.com:50051Source for API endpoint URL (opens in a new tab)
Languages
Korean, English, JapaneseSource for Languages (opens in a new tab)
Input audio format
Headerless PCM, 16 kHz, mono, 16 bits per sample.Source for Input audio format (opens in a new tab)
Session duration
Not collected

Both Korean and English error tables omit the numeric lifespan threshold. Do not infer a maximum session duration.

Word timestamps
SupportedSource for Word timestamps (opens in a new tab)

alignInfos includes start/end milliseconds and confidence for aligned units.

Endpoint pricing has not been collected. Check the provider for applicable rates.

Data & enterprise terms

Retention, training use, regional processing and enterprise terms have not been summarized for this endpoint.

API model ID
Not collected

This service exposes a dedicated route or workflow rather than a selectable checkpoint ID.

Access
CLOVA Speech domain and secret key required. Domain limits and permissions apply.
Regions
Korea; Classic and VPC environments are documented.
Request mode
Console-managed Object Storage batch workflow. Requires a Basic domain; one batch per domain, at most ten per account.Source for Request mode (opens in a new tab)
Audio duration limits
6 hours per fileSource for Audio duration limits (opens in a new tab)
Usage metering
Charges applied per input file; no discounted batch rate has been verified.Source for Usage metering (opens in a new tab)

Endpoint pricing has not been collected. Check the provider for applicable rates.

Data & enterprise terms

Retention, training use, regional processing and enterprise terms have not been summarized for this endpoint.

Caching

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.

CLOVA Speech

Not verified

Provider-side caching eligibility and rates have not been established for this service. A downloaded result or retained transcript is not prompt caching.

How it works
Not collected
Minimum reusable input
Not collected
Lifetime & refresh
Not collected
How to enable it
Not collected
Usage fields
Not collected
Billing rules
Not collected
Limits & data boundaries
Not collected

No verified cache rate for this endpoint. An empty rate is not a zero charge.

CLOVA Speech

Not verified

Provider-side caching eligibility and rates have not been established for this service. A downloaded result or retained transcript is not prompt caching.

How it works
Not collected
Minimum reusable input
Not collected
Lifetime & refresh
Not collected
How to enable it
Not collected
Usage fields
Not collected
Billing rules
Not collected
Limits & data boundaries
Not collected

No verified cache rate for this endpoint. An empty rate is not a zero charge.

CLOVA Speech

Not verified

Provider-side caching eligibility and rates have not been established for this service. A downloaded result or retained transcript is not prompt caching.

How it works
Not collected
Minimum reusable input
Not collected
Lifetime & refresh
Not collected
How to enable it
Not collected
Usage fields
Not collected
Billing rules
Not collected
Limits & data boundaries
Not collected

No verified cache rate for this endpoint. An empty rate is not a zero charge.

CLOVA Speech

Not verified

Provider-side caching eligibility and rates have not been established for this service. A downloaded result or retained transcript is not prompt caching.

How it works
Not collected
Minimum reusable input
Not collected
Lifetime & refresh
Not collected
How to enable it
Not collected
Usage fields
Not collected
Billing rules
Not collected
Limits & data boundaries
Not collected

No verified cache rate for this endpoint. An empty rate is not a zero charge.

CLOVA Speech

Not verified

Provider-side caching eligibility and rates have not been established for this service. A downloaded result or retained transcript is not prompt caching.

How it works
Not collected
Minimum reusable input
Not collected
Lifetime & refresh
Not collected
How to enable it
Not collected
Usage fields
Not collected
Billing rules
Not collected
Limits & data boundaries
Not collected

No verified cache rate for this endpoint. An empty rate is not a zero charge.

CLOVA Speech

Not verified

Provider-side caching eligibility and rates have not been established for this service. A downloaded result or retained transcript is not prompt caching.

How it works
Not collected
Minimum reusable input
Not collected
Lifetime & refresh
Not collected
How to enable it
Not collected
Usage fields
Not collected
Billing rules
Not collected
Limits & data boundaries
Not collected

No verified cache rate for this endpoint. An empty rate is not a zero charge.

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

Inferred for this named native commercial model service from NAVER’s own API documentation and the reviewed public HyperCLOVA X inventory. No weights for this exact service were identified. This is a service-level identity, not a claim that its underlying checkpoint is disclosed or that every NAVER audio model is proprietary. Public SEED Omni remains excluded.

Classification evidence (opens in a new tab)

Sources reviewed