Voicebox Multilingual · VB-Multi

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

Text, AudioSource 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

Not collected

Batch processing

Not collected

Fine-tuning

Not collected

Knowledge cutoff

Not collected

Release date

2023-06-16Source for Release date (opens in a new tab)

Research disclosure date, not general API or weight availability.

Parameters

330M acoustic model; 34M duration model; separate 13M HiFi-GAN vocoderSource for Parameters (opens in a new tab)

Component counts are explicitly scoped rather than reported as a rounded total.

Architecture

Non-autoregressive flow-matching audio infilling; 24-layer transformer with ALiBi and symmetric skip connectionsSource for Architecture (opens in a new tab)

Languages

English, French, German, Spanish, Polish, PortugueseSource for Languages (opens in a new tab)

Context caching

Not collected

No model-specific inference service or discounted prompt-cache contract is established for these research weights.

outputSampleRate

16 kHz waveform via HiFi-GANSource for outputSampleRate (opens in a new tab)

Published training data

50,000 hours of audiobooks across six languagesSource for Published training data (opens in a new tab)

speechTasks

Zero-shot text-to-speech, Style transfer, Transient noise removal, Spoken-content editing, Diverse speech samplingSource for speechTasks (opens in a new tab)

Prices belong to an endpoint and billing tier. A provider listing does not guarantee access for every account or region.

Caching can reuse earlier input processing. Availability, lifetime and billing depend on the endpoint. A cache hit still generates a new response.

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.

Why it is included

Meta explicitly states that Voicebox model weights and code are not being made public in its research release. This record identifies the original VB-En or VB-Multi model in the paper, not a third-party recreation or another product named Voicebox.

Lifecycle dates

Sources reviewed