Private zero-shot speech research model that uses phoneme alignment and merged audio codes to improve synthesis robustness and reduce autoregressive steps.
First arXiv disclosure of the research family. The reviewed paper may contain later revisions; this is not a commercial release date or a separate variant announcement date.
Architecture
Autoregressive and non-autoregressive Transformers, each with 12 layers, 16 heads, 1,024-dimensional embeddings and 4,096-dimensional feed-forward layers.Source for Architecture (opens in a new tab)
Read the paper and listen to prerecorded research samples. This is not an interactive model service and does not accept inference requests.
Regions
Not applicable to prerecorded research samples.
Endpoint pricing has not been collected. Check the provider for applicable rates.
Data & enterprise terms
No customer inference service or data-processing policy is established by the research demo.
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.
Microsoft Research (prerecorded samples)
Not verified
No public endpoint cache contract for this private research model.
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.
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
The original Microsoft research project explicitly states that it has no plans to provide public access or incorporate this model into a product. The reviewed original project and paper provide no full-model weight release. Open codec/vocoder components and third-party reproductions do not establish access to these original checkpoints.
