Narvatar
Narvatar is the name Nari gives its own human-focused avatar model. The company describes versions for streaming and non-streaming generation. Public model access is not confirmed in the reviewed materials.
Access not confirmed
The model appears in Nari's custom-training work
Nari presents Narvatar beside Dia within its custom-training and distillation work. That service description does not establish the avatar model's exact inputs or output limits.Nari's own speech and avatar research
Confirm the deliverable before planning an integration
Practical guidance. Ask for the supported input format, output resolution, identity controls and the available deployment package. A private demonstration, a hosted service and transferable model weights are different deliverables. The public description does not settle those choices.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Status | Access not confirmed |
| Published size description | Under five billion parameters. Nari does not publish an exact count on the reviewed pages.Nari's own speech and avatar research |
| Named variants | Streaming and non-streaming; separate checkpoint identifiers are not given.Nari's published model and service overview |
| Access evidence | Presented as in-house research on the company's training page. No download, model-specific pricing or public request schema is provided there.Nari's own speech and avatar research |
Related models
The larger Dia2 streaming dialogue model, with a wider decoder, audio-prefix conditioning and word-timing output.
Launch date not confirmedDownloadable