Skip to content
repository radar
Voice, Vision & Multimodal

🧬OmniGen2

ACTIVE STEADY

VectorSpaceLab/OmniGen2

OmniGen2: Exploration to Advanced Multimodal Generation. https://arxiv.org/abs/2506.18871

⭐ Popular: 4k stars, gaining about 2 a week

View on GitHub ↗

repo profile

vintage 2025 1 year old
language Jupyter Notebook
license Apache-2.0

momentum

total stars 4k
stars added last week +2
commits / week 0 steady
issues closed 15% ~7d to close, median
contributors 10
release cadence n/a
last activity 4mo ago

durability

backing foundation-backed academic (BAAI)
openness permissive

since we covered it

monthly average + 68/mo (+2%/mo) · + 912 total since PR#12

why it's a big deal

  • Combines four multimodal tasks in one model, visual understanding, text-to-image generation, instruction-guided image editing, and in-context generation across humans, objects, and scenes.
  • Ships under Apache-2.0 with weights on Hugging Face and ModelScope, training code, and the X2I2 training dataset, so researchers can reproduce and build on it rather than only calling an API.
  • Runs locally via a Gradio interface, a chat interface, or command-line scripts, giving practitioners a self-hostable open alternative for combined image editing and generation.

under the hood

  • Written in Python on PyTorch 2.6.0, with optional flash-attention for faster inference.
  • Uses two separate decoding pathways for text and image modalities with unshared parameters and a decoupled image tokenizer, which distinguishes it from OmniGen v1.
  • Inherits image interpretation from a Qwen-VL-2.5 foundation and exposes controls for text and image guidance scale, inference steps defaulting to 50, and Euler or DPMSolver++ schedulers.

Radar summary, generated from the project's public sources

star history

PR#12 · 3k4k now Jun 2025Aug 2026
  1. PR#12 3k 2025-07-09
  2. now 4k + 912 since first covered

curve is sampled from GitHub's star history, plus our own daily readings since we covered it; the dashed stretch is before we first covered it, the solid line since. figures at coverage are the numbers we printed then (approx.), current count is live.

understory

Better known than its recent output, coasting a little on attention.

-32 understory score output 0 · clout 32
Aug 2025 Jul 2026
  • output, commits & releases
  • clout, star velocity

output = commits & releases; clout = star velocity, both 0 to 100 monthly indices; the gap where output runs above clout is the understory. The understory →

covered in

  • PR#12 2025-07-09 below the radar

    Multimodal Generation Model for Images & Text

similar projects

compare these →
  • 🖼️ Immich

    TypeScript · 27× the stars · AGPL-3.0 licensed

    High performance self-hosted photo and video management solution.

    111k ACTIVE
  • 📡 RuView

    Rust · 22× the stars

    π RuView turns commodity WiFi signals into real-time spatial intelligence, vital sign monitoring, and presence detection, all without a single pixel of video.

    90k ACTIVE
  • 💡 PaddleOCR

    Python · 21× the stars

    Turn any PDF or image document into structured data for your AI. A powerful, lightweight OCR toolkit that bridges the gap between images/PDFs and LLMs. Supports 100+ languages.

    88k ACTIVE

comments

Sign in with GitHub to add your blip on OmniGen2.

loading comments…