Skip to content
Model infoModelsSnapshot: 15 Oct 2026

Gradium / Cloud TTS

Gradium TTS · September 2026

Current stable streaming speech synthesis release. Both default and the beta alias resolve to this release at the review date.

Speech & voice
availableClosed weightsText-to-speech
Official documentation (opens in a new tab)Reviewed 13 Oct 2026
15 of 25 specification fields collected
Text input limitFree tier: 1,500 characters per session. A paid-tier character cap is not established.See specification notes below
Languages5 listedFull language list in specifications
Input billingSee ratesGradium API
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
Maximum output
Not collected
Input modalities
TextSource for Input modalities (opens in a new tab)
Output modalities
AudioSource for Output modalities (opens in a new tab)
Reasoning
Not collected
Tool calling
Not collected
Structured output
Not collected
Streaming
SupportedSource for Streaming (opens in a new tab)
Batch processing
Not collected
Fine-tuning
SupportedSource for Fine-tuning (opens in a new tab)

Professional voice cloning is managed voice-specific fine-tuning, not unrestricted access to train the base model.

Knowledge cutoff
Not collected
Release date
2026-09-30Source for Release date (opens in a new tab)
Parameters
Not collected
Architecture
Not collected
Languages
English, French, German, Spanish, PortugueseSource for Languages (opens in a new tab)
Context caching
Not collected
Maximum prompt characters
Free tier: 1,500 characters per session. A paid-tier character cap is not established.Source for Maximum prompt characters (opens in a new tab)
Session duration
Up to 3,000 seconds per TTS session on WebSocket and RESTSource for Session duration (opens in a new tab)
Output formats
WAV, PCM, Ogg Opus, PCM 8/16/22.05/24/44.1/48 kHz, 8 kHz mu-law / A-lawSource for Output formats (opens in a new tab)
Audio sample rates
Default PCM/WAV: 48 kHz, 16-bit mono; PCM chunks contain 3,840 samples (80 ms)Source for Audio sample rates (opens in a new tab)
Generation controls
Temperature 0–1.4 and padding_bonus −4 to 4 are the tested ranges; rewrite rules and break/flush tags. cfg_coef is accepted but ignored.Source for Generation controls (opens in a new tab)

Validation allows wider ranges. Pronunciation dictionaries are WebSocket-only; REST supports rewrite_rules instead.

Voice cloning
Instant clone from about ten seconds. Pro voice clone requires at least 30 minutes; two hours recommended.Source for Voice cloning (opens in a new tab)
Published price
Standard TTS: 1 credit/character. Paid-plan overage rates vary.Source for Published price (opens in a new tab)
Custom model training
Pro voice clone: 1M credits for training plus 1.2 credits/character for synthesis. M includes five slots and L includes twenty; enterprise custom.Source for Custom model training (opens in a new tab)

Slot entitlements come from the plan table; usage and training credit costs come from its expanded FAQ.

Request quotas
TTS concurrency by Free / XS / S / M / L: 2 / 5 / 5 / 10 / 15; enterprise customSource for Request quotas (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.

Gradium API

Provider source (opens in a new tab)
API model ID
default
Access
Current stable TTS. gradium-tts-beta currently resolves to the same release and will follow future betas. API key and a voice ID required.
Regions
Default API routes to the nearest cluster without residency commitment. Paid enrolled organizations can pin EU or US inference; account data and custom voices remain in the EU, including for US inference.
API endpoint URL
wss://api.gradium.ai/api/speech/ttsSource for API endpoint URL (opens in a new tab)
Streaming
SupportedSource for Streaming (opens in a new tab)
Session duration
Up to 3,000 seconds per TTS session on WebSocket and RESTSource for Session duration (opens in a new tab)
Maximum prompt characters
Free tier: 1,500 characters per session. A paid-tier character cap is not established.Source for Maximum prompt characters (opens in a new tab)
Request mode
Bidirectional text/audio WebSocket; flush, reusable sessions and multiplexing.Source for Request mode (opens in a new tab)
Generation controls
pronunciation_id dictionaries and rewrite_rules supported.Source for Generation controls (opens in a new tab)

Prices retain their published currencies and billing units below.

Published billing units · original currency · observed 13 Oct 2026
CategoryUsageRateUnit
Speech usageSpeech synthesis · XS overage1$69.001M characters
SubscriptionPlan fee · XS2$13.00month
Speech usageSpeech synthesis · S overage3$50.001M characters
SubscriptionPlan fee · S4$43.00month
Speech usageSpeech synthesis · M overage5$40.001M characters
SubscriptionPlan fee · M6$340.00month
Speech usageSpeech synthesis · L overage7$38.001M characters
SubscriptionPlan fee · L8$1,615.00month
Voice trainingPro clone training · M9$40.00voice training
Speech usagePro clone synthesis · M9$48.001M characters
Voice trainingPro clone training · L10$38.00voice training
Speech usagePro clone synthesis · L10$45.601M characters

Rate conditions

  1. XS: $13/month includes 225,000 shared credits. Extra usage costs $6.9 per 100,000 credits and must be enabled. Included credits expire at month end; remaining pay-as-you-go credits roll over. Derived overage rate, not the effective rate of included credits. Annual billing and enterprise contracts are separate. Uses the documented credit multiplier; no text-token conversion.
  2. Includes 225,000 credits shared across speech products. Do not add the full subscription to every request. Monthly billing selected; annual pricing is a separate commitment.
  3. S: $43/month includes 900,000 shared credits. Extra usage costs $5 per 100,000 credits and must be enabled. Included credits expire at month end; remaining pay-as-you-go credits roll over. Derived overage rate, not the effective rate of included credits. Annual billing and enterprise contracts are separate. Uses the documented credit multiplier; no text-token conversion.
  4. Includes 900,000 credits shared across speech products. Do not add the full subscription to every request. Monthly billing selected; annual pricing is a separate commitment.
  5. M: $340/month includes 9,000,000 shared credits. Extra usage costs $4 per 100,000 credits and must be enabled. Included credits expire at month end; remaining pay-as-you-go credits roll over. Derived overage rate, not the effective rate of included credits. Annual billing and enterprise contracts are separate. Uses the documented credit multiplier; no text-token conversion.
  6. Includes 9,000,000 credits shared across speech products. Do not add the full subscription to every request. Monthly billing selected; annual pricing is a separate commitment.
  7. L: $1615/month includes 45,000,000 shared credits. Extra usage costs $3.8 per 100,000 credits and must be enabled. Included credits expire at month end; remaining pay-as-you-go credits roll over. Derived overage rate, not the effective rate of included credits. Annual billing and enterprise contracts are separate. Uses the documented credit multiplier; no text-token conversion.
  8. Includes 45,000,000 credits shared across speech products. Do not add the full subscription to every request. Monthly billing selected; annual pricing is a separate commitment.
  9. M paid-plan overage conversion at $4/100k credits. Pro voice slots: M (5), L (20); Enterprise by contract. Subscription includes shared credits; training is a one-time charge, not per synthesis.
  10. L paid-plan overage conversion at $3.8/100k credits. Pro voice slots: M (5), L (20); Enterprise by contract. Subscription includes shared credits; training is a one-time charge, not per synthesis.
Pricing source (opens in a new tab)Pricing source (opens in a new tab)Pricing source (opens in a new tab)

Data & enterprise terms

Paid plans can disable speech-session payload retention in organization Privacy Settings; this is opt-in, not a universal default. Standard ZDR excludes voice-design prompts and clone samples. Enterprise can contract extended coverage. Terms permit improvement/training use unless ZDR is activated; feedback has separate handling.

Read the service policy (opens in a new tab)

Gradium API

Provider source (opens in a new tab)
API model ID
default
Access
Current stable TTS. gradium-tts-beta currently resolves to the same release and will follow future betas. API key and a voice ID required.
Regions
Default API routes to the nearest cluster without residency commitment. Paid enrolled organizations can pin EU or US inference; account data and custom voices remain in the EU, including for US inference.
API endpoint URL
https://api.gradium.ai/api/post/speech/ttsSource for API endpoint URL (opens in a new tab)
Streaming
SupportedSource for Streaming (opens in a new tab)
Session duration
Up to 3,000 seconds per TTS session on WebSocket and RESTSource for Session duration (opens in a new tab)
Maximum prompt characters
Free tier: 1,500 characters per session. A paid-tier character cap is not established.Source for Maximum prompt characters (opens in a new tab)
Request mode
Whole text request; response can be raw audio or streaming JSON with audio and text timing.Source for Request mode (opens in a new tab)
Generation controls
rewrite_rules supported; pronunciation_id dictionaries are not available on REST.Source for Generation controls (opens in a new tab)

Prices retain their published currencies and billing units below.

Published billing units · original currency · observed 13 Oct 2026
CategoryUsageRateUnit
Speech usageSpeech synthesis · XS overage1$69.001M characters
SubscriptionPlan fee · XS2$13.00month
Speech usageSpeech synthesis · S overage3$50.001M characters
SubscriptionPlan fee · S4$43.00month
Speech usageSpeech synthesis · M overage5$40.001M characters
SubscriptionPlan fee · M6$340.00month
Speech usageSpeech synthesis · L overage7$38.001M characters
SubscriptionPlan fee · L8$1,615.00month
Voice trainingPro clone training · M9$40.00voice training
Speech usagePro clone synthesis · M9$48.001M characters
Voice trainingPro clone training · L10$38.00voice training
Speech usagePro clone synthesis · L10$45.601M characters

Rate conditions

  1. XS: $13/month includes 225,000 shared credits. Extra usage costs $6.9 per 100,000 credits and must be enabled. Included credits expire at month end; remaining pay-as-you-go credits roll over. Derived overage rate, not the effective rate of included credits. Annual billing and enterprise contracts are separate. Uses the documented credit multiplier; no text-token conversion.
  2. Includes 225,000 credits shared across speech products. Do not add the full subscription to every request. Monthly billing selected; annual pricing is a separate commitment.
  3. S: $43/month includes 900,000 shared credits. Extra usage costs $5 per 100,000 credits and must be enabled. Included credits expire at month end; remaining pay-as-you-go credits roll over. Derived overage rate, not the effective rate of included credits. Annual billing and enterprise contracts are separate. Uses the documented credit multiplier; no text-token conversion.
  4. Includes 900,000 credits shared across speech products. Do not add the full subscription to every request. Monthly billing selected; annual pricing is a separate commitment.
  5. M: $340/month includes 9,000,000 shared credits. Extra usage costs $4 per 100,000 credits and must be enabled. Included credits expire at month end; remaining pay-as-you-go credits roll over. Derived overage rate, not the effective rate of included credits. Annual billing and enterprise contracts are separate. Uses the documented credit multiplier; no text-token conversion.
  6. Includes 9,000,000 credits shared across speech products. Do not add the full subscription to every request. Monthly billing selected; annual pricing is a separate commitment.
  7. L: $1615/month includes 45,000,000 shared credits. Extra usage costs $3.8 per 100,000 credits and must be enabled. Included credits expire at month end; remaining pay-as-you-go credits roll over. Derived overage rate, not the effective rate of included credits. Annual billing and enterprise contracts are separate. Uses the documented credit multiplier; no text-token conversion.
  8. Includes 45,000,000 credits shared across speech products. Do not add the full subscription to every request. Monthly billing selected; annual pricing is a separate commitment.
  9. M paid-plan overage conversion at $4/100k credits. Pro voice slots: M (5), L (20); Enterprise by contract. Subscription includes shared credits; training is a one-time charge, not per synthesis.
  10. L paid-plan overage conversion at $3.8/100k credits. Pro voice slots: M (5), L (20); Enterprise by contract. Subscription includes shared credits; training is a one-time charge, not per synthesis.
Pricing source (opens in a new tab)Pricing source (opens in a new tab)Pricing source (opens in a new tab)

Data & enterprise terms

Paid plans can disable speech-session payload retention in organization Privacy Settings; this is opt-in, not a universal default. Standard ZDR excludes voice-design prompts and clone samples. Enterprise can contract extended coverage. Terms permit improvement/training use unless ZDR is activated; feedback has separate handling.

Read the service policy (opens in a new tab)

LiveKit Inference

Provider source (opens in a new tab)
API model ID
gradium/default
Access
LiveKit Inference hosts the moving Gradium default; a pinned September checkpoint is not promised. Flagship voices supported; pre-existing custom/community voices require the Gradium plugin and a native account.
Regions
Global endpoint and dedicated EU endpoint documented.
Streaming
SupportedSource for Streaming (opens in a new tab)
Languages
English, French, German, Spanish, PortugueseSource for Languages (opens in a new tab)

Prices retain their published currencies and billing units below.

Published billing units · original currency · observed 13 Oct 2026
CategoryUsageRateUnit
Speech usageSpeech synthesis · Build / Ship1$48.001M characters
Speech usageSpeech synthesis · Scale1$36.001M characters

Rate conditions

  1. LiveKit hosted rate; plan fees, included credits and enterprise terms are separate. Not the native Gradium credit price.
Pricing source (opens in a new tab)

Data & enterprise terms

LiveKit Inference documents no content retention by default, with observability and custom-voice handling governed separately. Native account settings are not automatically inherited.

Read the service policy (opens in a new tab)

Baseten

Provider source (opens in a new tab)
API model ID
Not collected

A pinned checkpoint selector is not established.

Access
Dedicated deployment by arrangement. The guide does not pin its checkpoint to the September cloud release; verify the contracted image. Uses the supplied WebSocket URL directly and a Baseten API key.
Regions
Dedicated deployment region depends on the contract.
Streaming
SupportedSource for Streaming (opens in a new tab)
Voice references
346 voices plus a built-in default in the documented deployment imageSource for Voice references (opens in a new tab)
Generation controls
Inline rewrites; hosted pronunciation_id dictionaries are unavailable.Source for Generation controls (opens in a new tab)

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

Data & enterprise terms

Dedicated deployment guide does not establish a universal retention/training policy. Confirm the deployment contract; no native ZDR guarantee is transferred.

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.

Gradium API

default
Not verified

No reviewed model-specific prompt-cache contract. Reused WebSockets, stored voices and pronunciation dictionaries do not establish discounted prompt caching.

How it works
Not collected

No reviewed model-specific prompt-cache contract. Reused WebSockets, stored voices and pronunciation dictionaries do not establish discounted prompt caching.

Minimum reusable input
Not collected

No reviewed model-specific prompt-cache contract. Reused WebSockets, stored voices and pronunciation dictionaries do not establish discounted prompt caching.

Lifetime & refresh
Not collected

No reviewed model-specific prompt-cache contract. Reused WebSockets, stored voices and pronunciation dictionaries do not establish discounted prompt caching.

How to enable it
Not collected

No reviewed model-specific prompt-cache contract. Reused WebSockets, stored voices and pronunciation dictionaries do not establish discounted prompt caching.

Usage fields
Not collected

No reviewed model-specific prompt-cache contract. Reused WebSockets, stored voices and pronunciation dictionaries do not establish discounted prompt caching.

Billing rules
Not collected

No reviewed model-specific prompt-cache contract. Reused WebSockets, stored voices and pronunciation dictionaries do not establish discounted prompt caching.

Limits & data boundaries
Not collected

No reviewed model-specific prompt-cache contract. Reused WebSockets, stored voices and pronunciation dictionaries do not establish discounted prompt caching.

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

Gradium API

default
Not verified

No reviewed model-specific prompt-cache contract. Reused WebSockets, stored voices and pronunciation dictionaries do not establish discounted prompt caching.

How it works
Not collected

No reviewed model-specific prompt-cache contract. Reused WebSockets, stored voices and pronunciation dictionaries do not establish discounted prompt caching.

Minimum reusable input
Not collected

No reviewed model-specific prompt-cache contract. Reused WebSockets, stored voices and pronunciation dictionaries do not establish discounted prompt caching.

Lifetime & refresh
Not collected

No reviewed model-specific prompt-cache contract. Reused WebSockets, stored voices and pronunciation dictionaries do not establish discounted prompt caching.

How to enable it
Not collected

No reviewed model-specific prompt-cache contract. Reused WebSockets, stored voices and pronunciation dictionaries do not establish discounted prompt caching.

Usage fields
Not collected

No reviewed model-specific prompt-cache contract. Reused WebSockets, stored voices and pronunciation dictionaries do not establish discounted prompt caching.

Billing rules
Not collected

No reviewed model-specific prompt-cache contract. Reused WebSockets, stored voices and pronunciation dictionaries do not establish discounted prompt caching.

Limits & data boundaries
Not collected

No reviewed model-specific prompt-cache contract. Reused WebSockets, stored voices and pronunciation dictionaries do not establish discounted prompt caching.

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

LiveKit Inference

gradium/default
Not verified

No reviewed model-specific prompt-cache contract. Reused WebSockets, stored voices and pronunciation dictionaries do not establish discounted prompt caching.

How it works
Not collected

No reviewed model-specific prompt-cache contract. Reused WebSockets, stored voices and pronunciation dictionaries do not establish discounted prompt caching.

Minimum reusable input
Not collected

No reviewed model-specific prompt-cache contract. Reused WebSockets, stored voices and pronunciation dictionaries do not establish discounted prompt caching.

Lifetime & refresh
Not collected

No reviewed model-specific prompt-cache contract. Reused WebSockets, stored voices and pronunciation dictionaries do not establish discounted prompt caching.

How to enable it
Not collected

No reviewed model-specific prompt-cache contract. Reused WebSockets, stored voices and pronunciation dictionaries do not establish discounted prompt caching.

Usage fields
Not collected

No reviewed model-specific prompt-cache contract. Reused WebSockets, stored voices and pronunciation dictionaries do not establish discounted prompt caching.

Billing rules
Not collected

No reviewed model-specific prompt-cache contract. Reused WebSockets, stored voices and pronunciation dictionaries do not establish discounted prompt caching.

Limits & data boundaries
Not collected

No reviewed model-specific prompt-cache contract. Reused WebSockets, stored voices and pronunciation dictionaries do not establish discounted prompt caching.

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

Baseten

Not verified

No reviewed model-specific prompt-cache contract. Reused WebSockets, stored voices and pronunciation dictionaries do not establish discounted prompt caching.

How it works
Not collected

No reviewed model-specific prompt-cache contract. Reused WebSockets, stored voices and pronunciation dictionaries do not establish discounted prompt caching.

Minimum reusable input
Not collected

No reviewed model-specific prompt-cache contract. Reused WebSockets, stored voices and pronunciation dictionaries do not establish discounted prompt caching.

Lifetime & refresh
Not collected

No reviewed model-specific prompt-cache contract. Reused WebSockets, stored voices and pronunciation dictionaries do not establish discounted prompt caching.

How to enable it
Not collected

No reviewed model-specific prompt-cache contract. Reused WebSockets, stored voices and pronunciation dictionaries do not establish discounted prompt caching.

Usage fields
Not collected

No reviewed model-specific prompt-cache contract. Reused WebSockets, stored voices and pronunciation dictionaries do not establish discounted prompt caching.

Billing rules
Not collected

No reviewed model-specific prompt-cache contract. Reused WebSockets, stored voices and pronunciation dictionaries do not establish discounted prompt caching.

Limits & data boundaries
Not collected

No reviewed model-specific prompt-cache contract. Reused WebSockets, stored voices and pronunciation dictionaries do not establish discounted prompt caching.

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

Voice choices

71 voices

Choose a voice within Gradium TTS · September 2026. These are API voice identifiers, not additional model releases. Source links document each voice.

Published voice catalog
Voice / API IDLanguage & accentDelivery
ZoeyNbpkqMVS3CJeq2j8English
United States (US)		Playful, upbeat and Gen Z energy voice with a standard American accent. Perfect for engaging conversations! Curated flagship voice, not a separate model.Voice source (opens in a new tab)
SunnieYVzbrdWnnu9FgRn5English
United States (US)		A joyful young American voice at medium-high pitch, warm and bubbly, keeping every support call bright and light. Curated flagship voice, not a separate model.Voice source (opens in a new tab)
MarloweBla6SbVMczYnOhfKEnglish
United States (US)		A bubbly, warm young American voice full of girly charm and a laugh under every word, ideal for friendly support and assistant chats. Curated flagship voice, not a separate model.Voice source (opens in a new tab)
Harper4SZHfMpw-p46YwgsEnglish
United States (US)		Modern, confident and friendly voice with a standard American accent. Curated flagship voice, not a separate model.Voice source (opens in a new tab)
BrooklynD6COLz20Hw7uh3UKEnglish
United States (US)		A warm, effusive young American voice with girly charm and contagious laughter, greeting customers and making everyone feel at home. Curated flagship voice, not a separate model.Voice source (opens in a new tab)
Sterling6MFfc37kq0sBjBjyEnglish
United States (US)		A warm energetic American adult voice with theatrical flair that makes every sentence feel like the start of something big. Curated flagship voice, not a separate model.Voice source (opens in a new tab)
Russell_6Aslh2DxfmnRLmPEnglish
United States (US)		A high-energy American adult voice that pushes and encourages with the intensity of someone who genuinely believes in you. Curated flagship voice, not a separate model.Voice source (opens in a new tab)
Marcusr2sIQdqqoqgRJuXwEnglish
United States (US)		A high-energy resonant American adult voice that speaks with the unshakeable conviction of someone who's already sold you. Curated flagship voice, not a separate model.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

The proprietary service license reserves ownership of models and restricts copying, redistribution and extraction. Phonon is supplied as a separately licensed, partner-specific binary; offline deployment is not an open-weight license. Public SDKs and evaluation datasets do not grant production model weights.

Classification evidence (opens in a new tab)

Lifecycle dates

Release documented30 Sept 2026Source (opens in a new tab)
Public beta announced22 Sept 2026Source (opens in a new tab)

Sources reviewed