vllm-project/vllm-omni

[New Model]: Sana-WM

Open

#3,656 opened on May 16, 2026

View on GitHub
 (6 comments) (2 reactions) (1 assignee)Python (1,067 forks)github user discovery
help wantednew model

Repository metrics

Stars
 (4,990 stars)
PR merge metrics
 (PR metrics pending)

Description

The model to consider.

a 2.6B camera-controlled, 720p, 1 minute-long video generation model, ICLR 2026 from Nvidia

Code: https://github.com/NVlabs/Sana Website: https://nvlabs.github.io/Sana/WM/

cc: @TKONIY Related to: https://github.com/vllm-project/vllm-omni/issues/1987

The closest model vllm-omni already supports.

LTX-2.

What's your difficulty of supporting the model you want?

It needs a text encoder (Gemma), a DiT backbone, a camera control (dual branch) module, a LTX2 video VAE tokenizer.

The new parts needed: for camera control, need to create a new processor in stage_input_processors. Also need to add a (Gated DeltaNet) attention backend in diffusion/attention/backends/.

The parts we already have: LTX2 video tokenizer, DiT, gemma text encoder

Use case and motivation

No response

Before submitting a new issue...

  • Make sure you already searched for relevant issues, and asked the chatbot living at the bottom right corner of the documentation page, which can answer lots of frequently asked questions.

Contributor guide

Research direction
Familiarize yourself with the existing model structure (e.g., LTX 2) and the stage input processors and attention backends. Implement a new processor for camera control and a Gated DeltaNet attention backend, then integrate with the existing DiT backbone and Gemma text encoder.
Tech stack
pythonpytorch
Domain
backendaimachine learning
Issue type
Feature
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
3
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
1-2 days
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
Active
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
Clear
Prerequisites
GitPython
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
30