LibreAI / Image / OmniGen
// image · VectorSpaceLab/OmniGen

OmniGen

OmniGen est un projet de image open source, créé par VectorSpaceLab (VectorSpaceLab/OmniGen sur GitHub). Il compte 4 343 étoiles sur GitHub et 364 forks, il est publié sous la licence MIT et a été mis à jour pour la dernière fois sur GitHub le 4 décembre 2025.

OmniGen: one unified model for text-to-image, editing, subject-driven and identity-preserving generation

★ 4,3 k étoiles⑂ 364 forksMITmis à jour déc. 2025
▦ visuel — le README de ce projet ne contient aucune capture intégrable,
voici donc un emplacement réservé. Le guide ci-dessous est complet malgré tout.
// détails du projet

Détails du projet

Étoiles GitHub
4 343
Forks
364
Licence
MIT
Dernière mise à jour sur GitHub
4 décembre 2025
Catégorie
Image
Type de fiche
Guide d'installation complet
// fonctionnement

Comment ça marche

  1. Prepare a GPU environment

    Create the documented Conda environment (Python 3.10.x) — OmniGen inference expects a CUDA GPU.

  2. Clone and install

    Clone the repo and install it in editable mode after installing the pinned PyTorch build.

  3. Load the model

    Load Shitao/OmniGen-v1 weights from Hugging Face through the OmniGen pipeline (or Diffusers).

  4. Generate with references

    Write a prompt, pass input images with guidance scales, and save your first edited or subject-driven image.

// fonctionnalités

Pourquoi on l'adore

Unified task coverage

Text-to-image, editing, subject-driven, identity-preserving, and image-conditioned generation in one model.

No extra plugins

Tasks that usually need ControlNet or IP-Adapter are handled natively by the framework.

In-prompt image references

Reference images inline with <|image_1|>-style placeholders and combine multiple inputs.

Guidance controls

Separate guidance_scale and img_guidance_scale parameters balance text and image influence.

Diffusers and Gradio paths

Use the model through Hugging Face Diffusers or run the included Gradio demo (app.py).

Fine-tuning included

LoRA and full-parameter training scripts (train.py with accelerate) adapt the model to your data.

Advertisement
// lancez-le

Lancez-le vous-même

  1. Clone the repository

    Clone the project and enter its directory.

    git clone https://github.com/VectorSpaceLab/OmniGen.git
    cd OmniGen
  2. Set up the environment

    Create the Conda env, install the pinned CUDA PyTorch build from the README, then install the package.

    conda create -n omnigen python=3.10.13
    conda activate omnigen
    pip install torch==2.3.1+cu118 torchvision --extra-index-url https://download.pytorch.org/whl/cu118
    pip install -e .
  3. Run inference

    Load the pretrained model and generate with an image reference, following the README's examples.

    from OmniGen import OmniGenPipeline
    pipe = OmniGenPipeline.from_pretrained("Shitao/OmniGen-v1")
    images = pipe(prompt="Edit <|image_1|>", input_images=["input.png"], guidance_scale=2.5, img_guidance_scale=1.6, seed=0)
// cas d'usage

Ce que vous pouvez créer

  • Edit a photo with a text instruction while preserving the subject's identity
  • Generate new images of a specific product or person from reference photos
  • Replace a ControlNet/IP-Adapter pipeline with a single unified model
  • Fine-tune a LoRA for a house illustration style
Advertisement
// faq

Questions fréquentes

Who maintains OmniGen?

VectorSpaceLab, the research team behind the OmniGen paper (arXiv 2409.11340).

Is OmniGen free?

The code is MIT-licensed and the weights are downloadable from Hugging Face; check the model card for the weights' own terms.

Does it need a paid API key?

No — it runs locally on your GPU with downloaded weights.

How does it compare to Stable Diffusion plus ControlNet?

OmniGen folds editing, subject-driven, and conditioned generation into one model with in-prompt image references, instead of a base model plus separate plugin networks.

How do I get started?

Create the Python 3.10 Conda env, install the pinned PyTorch build, `pip install -e .`, then run the OmniGenPipeline example from the README.

// continuez l'exploration

Outils similaires

Image

Bagel

ByteDance's open-source unified multimodal model for understanding, generation and image editing.

★ 6,2 k étoilesApache-2.0
Voir le guide →
Image

omnigen-vault

Infinite text-free AI image generator with a searchable multilingual vault

★ 35 étoilesMIT
Voir le guide →
Image

mmagic

MMagic — OpenMMLab's generative-AI toolbox: diffusion, text-to-image, restoration and a model zoo.

★ 7,5 k étoilesApache-2.0
Voir le guide →
Image

HunyuanImage-3.0

Tencent's HunyuanImage-3.0 — a powerful native multimodal model for image generation.

★ 3,3 k étoilesNOASSERTION
Voir le guide →
Image

pydantic-ai

How Python does AI. Agents, realtime voice, image generation, embeddings. Every model, every interface, typed end to end.

★ 20,4 k étoilesMIT
Voir le guide →
Image

zenith-image-generator

Multi-provider AI image generator with OpenAI-compatible API and one-click deploy

★ 223 étoilesCustom
Voir le guide →
// alternatives

Alternatives en Image

Les autres projets de la catégorie Image avec le plus d'étoiles du catalogue LibreAI — la sélection réaliste d'alternatives si OmniGen ne vous convient pas.

  • Stable Diffusion Webui★ 165,2 k

    Stable Diffusion web UI

  • ComfyUI★ 136,0 k

    The node-graph powerhouse for precise, professional AI image and video generation.

  • Immich★ 115,5 k

    High performance self-hosted photo and video management solution

  • PhotoCraft★ 73,2 k

    PhotoCraft — a clean-room Photoshop reimplementation in pure Rust: layers, masks, adjustment layers, type, vectors, brushes and real PSD files, offline, MIT / Apache-2.0

  • Fooocus★ 53,3 k

    Focus on prompting and generating

// plus en image

Plus en Image

Projets de la catégorie Image proches dans le catalogue, les plus proches en étoiles de OmniGen — continuez à explorer de part et d'autre.

  • lite.ai.toolkit★ 4,4 k

    A lite C++ AI toolkit: 100+ models with MNN, ORT and TRT, including Det, Seg, Stable-Diffusion, Face-Fusion.

  • BiRefNet★ 4,2 k

    [CAAI AIR'24] Bilateral Reference for High-Resolution Dichotomous Image Segmentation

  • SwarmUI★ 4,6 k

    SwarmUI (formerly StableSwarmUI), A Modular Stable Diffusion Web-User-Interface, with an emphasis on making powertools easily accessible, high performance, a…

  • RawTherapee★ 4,2 k

    A powerful cross-platform raw photo processing program

  • geekai★ 4,7 k

    AI 助手全套开源解决方案,自带运营管理后台,开箱即用。集成了 ChatGPT, Azure, ChatGLM,讯飞星火,文心一言等多个平台的大语言模型。支持 MJ AI 绘画,Stable Diffusion AI 绘画,微博热搜等插件工具。采用 Go + Vue3 + element-plus 实现。

  • Pinta★ 4,1 k

    Simple GTK Paint Program

Obtenir OmniGen

Gratuit et ouvert : récupérez le code source, mettez une étoile au dépôt et commencez à créer.