Skip to content
Model infoModelsSnapshot: 15 Oct 2026

HeyGen / HeyGen voice engines

Starfish

HeyGen catalog speech engine with Asian-language pronunciation and pacing specialization.

Speech & voice
availableClosed weightsText-to-speechSpeech synthesis
Official documentation (opens in a new tab)Reviewed 14 Oct 2026
5 of 18 specification fields collected
Text input limit5K5,000
LanguagesNot collectedAwaiting source review
Input billingSee ratesHeyGen
Output billingSee ratesPublished units and conditions below

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

No exact model-specific contract established; missing information does not mean unsupported.

Maximum output
Not collected

No exact model-specific contract established; missing information does not mean unsupported.

Context caching
SupportedSource for Context caching (opens in a new tab)

Generated speech reuse, not LLM prompt-token caching.

Reasoning
Not collected

No exact model-specific contract established; missing information does not mean unsupported.

Tool calling
Not collected

No exact model-specific contract established; missing information does not mean unsupported.

Structured output
Not collected

No exact model-specific contract established; missing information does not mean unsupported.

Streaming
Not collected

No exact model-specific contract established; missing information does not mean unsupported.

Batch processing
Not collected

No exact model-specific contract established; missing information does not mean unsupported.

Fine-tuning
Not collected

No exact model-specific contract established; missing information does not mean unsupported.

Parameters
Not collected

No exact model-specific contract established; missing information does not mean unsupported.

Architecture
Not collected

No exact model-specific contract established; missing information does not mean unsupported.

Knowledge cutoff
Not collected

No exact model-specific contract established; missing information does not mean unsupported.

Languages
Not collected

No exact model-specific contract established; missing information does not mean unsupported.

Release date
Not collected

No exact model-specific contract established; missing information does not mean unsupported.

Input modalities
TextSource for Input modalities (opens in a new tab)
Output modalities
AudioSource for Output modalities (opens in a new tab)
Maximum prompt characters
5,000Source for Maximum prompt characters (opens in a new tab)
Generation controls
Speed 0.5–2, language and locale controls, supported break pausesSource for Generation controls (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.

HeyGen

Provider source (opens in a new tab)
API model ID
starfish

Catalog engine selector, not a versioned checkpoint.

Access
POST https://api.heygen.com/v3/voices/speech with engine=starfish and an eligible catalog voice.
Regions
No model-specific processing or storage region guarantee established.
Languages
Asian-language specialization, including Chinese, Japanese and Korean; exact supported locales depend on the catalog voice.Source for Languages (opens in a new tab)
engineEligibility
Check available_engines and default_engine on the catalog voice. Unsupported combinations return 400. Starfish is the fallback when no saved engine is present.Source for engineEligibility (opens in a new tab)
pricingSource
Public API help-center table reviewed October 14, 2026. API and Studio plans are separate; pay-as-you-go credits expire after 12 months.Source for pricingSource (opens in a new tab)

Prices retain their published currencies and billing units below.

Published billing units · original currency · observed 14 Oct 2026
CategoryUsageRateUnit
Speech generationSpeech1$0.12minute of generated speech

Rate conditions

  1. Public self-serve API Starfish tariff. Do not apply to Voice 1, Turbo or Studio exports. Native minute pricing, not text tokens.
Pricing source (opens in a new tab)

Data & enterprise terms

Privacy policy distinguishes customer-directed processing from HeyGen controller purposes. Applicable plan terms, DPA and Enterprise agreements govern training use. No blanket zero-retention or no-training promise inferred. Deleted information can remain in disaster-recovery backups for 60 days.

Read the service policy (opens in a new tab)

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.

HeyGen

starfish
Documented
Support evidence (opens in a new tab)
How it works
Generated speech output cacheSource for How it works (opens in a new tab)
Minimum reusable input
Not collected

No model-specific prompt-cache contract established. Idempotency and saved avatar/voice assets are separate.

Lifetime & refresh
Not collected

No model-specific prompt-cache contract established. Idempotency and saved avatar/voice assets are separate.

How to enable it
Not collected

No model-specific prompt-cache contract established. Idempotency and saved avatar/voice assets are separate.

Usage fields
Set force_regenerate=true to skip the speech cache and synthesize fresh audio.Source for Usage fields (opens in a new tab)
Billing rules
Not collected

No model-specific prompt-cache contract established. Idempotency and saved avatar/voice assets are separate.

Limits & data boundaries
Only eligible catalog voices and supported engine/language combinations. No automatic fallback for an explicitly requested unavailable engine.Source for Limits & data boundaries (opens in a new tab)

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

Caching reviewed 14 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

Official documentation identifies this HeyGen model or proprietary engine, exposed through hosted API or Studio access. Service terms restrict distribution of proprietary technology; no model-weight distribution is established.

Classification evidence (opens in a new tab)

Sources reviewed