OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Image & Video AI

Omnilottie

OmniLottie is an open-source multimodal generative model that converts text, images and videos into vector animations encoded as Lottie JSON. Its repository provides inference code and model weights, with a Hugging Face demo, model-format support and a community-contributed ComfyUI integration.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Omnilottie — OpenViST project profile in Image & Video AI.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use OmniLottie when the desired deliverable is vector animation data in Lottie JSON and the prompt can be expressed through text, an image, a video, or a combination of inputs.

Who it is for

  • Developers adding instruction-guided Lottie generation to creative software or workflows.
  • Designers and animators who want to generate vector-animation data from text or visual references.
  • Researchers evaluating multimodal vector-animation generation.

Documented uses

  • Generate a Lottie animation from a text prompt.
  • Create an animation from an image and an accompanying text instruction.
  • Convert a video reference into Lottie JSON.
  • Evaluate vector-animation generation using the released MMLottieBench dataset.
Before you choose

Requirements & limits

The repository links to a Hugging Face-hosted Gradio demo and provides model weights and inference code for local use. Its documented self-hosted setup uses Python 3.10, PyTorch 2.3.0 with CUDA 12.1 support, and reports 15.2 GB GPU memory usage for inference.

  • The output is Lottie JSON, not a rendered video file.
  • The model accepts text, image and video inputs; the documented image example also includes a text instruction.
  • The README gives a substantial GPU-memory figure of 15.2 GB for inference.
  • Hugging Face-format and original-format weights use different inference scripts.
  • The project lists inference code and weights as released, while training code remains unreleased.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Run a text-to-Lottie inference with Hugging Face weights

Generate a Lottie JSON file from a text instruction using the documented Hugging Face inference script.

Before you start

  • Repository cloned locally with its dependencies installed.
  • Python 3.10 environment; the README's inference setup uses PyTorch 2.3.0 with CUDA 12.1 support.
  • OmniLottie Hugging Face model weights, either downloaded locally or fetched through the Hugging Face Hub.
  • The README reports 15.2 GB GPU memory usage for inference.

Steps

  1. Install the repository dependencies using its requirements file in the prepared environment.

  2. Download the Hugging Face model weights with the Hugging Face CLI, or use the model identifier directly.

  3. Run inference_hf.py with the model path, a text prompt and an output JSON filename.

Expected outcome

The inference script is configured to write generated animation data to the specified JSON output file.

Guide limitations

  • The repository documents inference commands but this guide does not independently execute them.
  • The README reports GPU-memory use; hardware availability and performance depend on the user's environment.
Documentation 1 ↗
Source-based practical guide

Generate Lottie JSON from an image or video

Use the Hugging Face inference script with a visual input, optionally adding a text instruction for an image.

Before you start

  • The repository's Python inference environment and Hugging Face-format weights.
  • An image file for image-to-Lottie or a video file for video-to-Lottie.
  • The README reports 15.2 GB GPU memory usage for inference.

Steps

  1. Prepare the image or video file and make the model weights available locally or through the Hugging Face Hub.

  2. For an image, pass the image path and a text instruction to inference_hf.py.

  3. For a video, pass the video path to inference_hf.py.

  4. Set the output argument to the desired JSON filename.

Expected outcome

The documented inference commands produce Lottie JSON from an image with an instruction or from a video.

Guide limitations

  • The image example includes a text instruction; the video example does not show one.
  • The commands are documented usage examples, not independently executed tests.
Documentation 1 ↗
COMPARE OPTIONS

Compare options for Model inference and serving

These projects document a common task and share a project role. This is a source-based comparison, not an execution benchmark or a guarantee of interchangeability.

ProjectDocumented purposePlatformAccess and requirements
OmnilottieGenerate a Lottie animation from a text prompt.Linux, Self-hostedThe repository links to a Hugging Face-hosted Gradio demo and provides model weights and inference code for local use. Its documented self-hosted setup uses Python 3.10, PyTorch 2.3.0 with CUDA 12.1 support, and reports 15.2 GB GPU memory usage for inference.
FASHN VTON v1.5Generate a try-on preview from a person image and a garment image.CLI, Linux, macOS, Self-hostedThe repository is public and states an Apache-2.0 license. Local setup requires downloading approximately 2 GB of model weights.
PosteromniRescale a poster to a target aspect ratio using a text prompt.
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
16 October 2026
Original website title
OpenVGLab/OmniLottie · Official homepage title · retained in its source language
How we review sourcesSuggest a correction