Skip to content
Model infoModelsSnapshot: 15 Oct 2026

Hyper3D / Deemos / Media2Face / GNPFA

GNPFA Geometry VAE

Research facial-expression model that separates expression codes from identity and reconstructs facial geometry.

3D processing
restrictedClosed weights3D processing
Official documentation (opens in a new tab)Reviewed 14 Oct 2026
6 of 21 specification fields collected
Context windowNot collectedAwaiting source review
Maximum outputNot collectedAwaiting source review
Input billingNot collectedMedia2Face research publication
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 input tokens
Not collected
Maximum output
Not collected
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
Not collected

Paper submission and revision dates are not weight-release dates.

Parameters
Not collected
Languages
Not collected
Context caching
Not collected
Architecture
Geometry encoder and neutral-geometry-conditioned U-Net decoder; UV coordinate maps share a fixed mesh topology.Source for Architecture (opens in a new tab)
Input modalities
Expressive geometry, Neutral geometrySource for Input modalities (opens in a new tab)
Output modalities
Expression latent, Reconstructed facial geometrySource for Output modalities (opens in a new tab)
Published training data
43,652 registered meshes and 698,432 images from 300 identities; synthetic expression augmentation uses personalized blendshapes for 200 identities.Source for Published training data (opens in a new tab)
Training hardware
Geometry VAE training: approximately ten days on NVIDIA A6000 with AdaBelief.Source for Training hardware (opens in a new tab)
Endpoint availability
Research edition; code and dataset are listed as upcoming by the official project.Source for Endpoint availability (opens in a new tab)
Usage metering
Not collected

No current hosted API price or callable checkpoint selector verified.

Access & pricing

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

Media2Face research publication

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

No public native selector for this research edition verified.

Access
Paper and project page only; no original weight download or current callable API verified.
Regions
Not established

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

Data & enterprise terms

No commercial retention or training-use contract assigned to a research publication.

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.

Media2Face research publication

Not verified

No model-specific caching contract verified.

How it works
Not collected

No model-specific caching contract verified.

Minimum reusable input
Not collected

No model-specific caching contract verified.

Lifetime & refresh
Not collected

No model-specific caching contract verified.

How to enable it
Not collected

No model-specific caching contract verified.

Usage fields
Not collected

No model-specific caching contract verified.

Billing rules
Not collected

No model-specific caching contract verified.

Limits & data boundaries
Not collected

No model-specific caching contract verified.

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

Qualified unreleased-weight research classification: the official project labels code and data upcoming, and its links lead back to the project. No original weight release was established in reviewed sources. This is not a claim of permanent proprietary licensing; publication alone does not release weights.

Classification evidence (opens in a new tab)

Lifecycle dates

Research paper first submitted28 Jan 2024Source (opens in a new tab)
Research paper revised30 Jan 2024Source (opens in a new tab)

Sources reviewed