Vimoo-1.0-Lite-MLX
Vimoo-1.0-Lite-MLX runs Vimoo-1.0-Lite TI2V Distilled on Apple Silicon with Vimoo-1.0-MLX. It contains the ≈9.5B Vimoo-1.0-Lite TI2V model, which handles both text-to-video (T2V) and image-to-video (I2V), with its distilled LoRA adapter and scheduler.
See the Vimoo 1.0 repository for the model family, online API, system overview, architecture and evaluation.
🔥 News
- [2026-10] Release Vimoo 1.0: T2V / I2V models, Diffusers inference examples and ComfyUI workflows.
🎬 Showcase
🚀 Quick Start
Installation
Use macOS 14 or newer on Apple Silicon, native arm64 Python 3.11 or newer, and ffmpeg for MP4 output. The installation includes MLX 0.32.2. Install the MLX runtime:
git clone https://github.com/Vimoo-AI/Vimoo-1.0-MLX.git
cd Vimoo-1.0-MLX
python3 -m venv .venv
source .venv/bin/activate
python3 -m pip install -e .
See the installation guide for environment requirements.
Download weights
hf download Vimoo-AI/Vimoo-1.0-Lite-MLX --local-dir ./models/Vimoo-1.0-Lite-MLX
Shared components are downloaded at the revisions recorded in modular_model_index.json. The loader fetches only the required files. Set HF_HOME to choose the download cache; after downloading, HF_HUB_OFFLINE=1 allows offline use.
Generate video
Replace the image with your own file:
vimoo-mlx --model ./models/Vimoo-1.0-Lite-MLX --prompt 'A young woman in a red coat walks through a snowy city street and looks up at the falling snow.' --output lite-t2v.mp4
vimoo-mlx --model ./models/Vimoo-1.0-Lite-MLX --prompt 'The dog shakes its whole body, then sits back down in the same spot. The camera remains fixed throughout.' --image dog.png --output lite-i2v.mp4
Supplying --image selects image-to-video and uses the image as the first frame; omitting it selects text-to-video at 1280 × 720. Both tasks default to 121 frames at 24 fps with the included twelve-step distilled recipe. See the Vimoo-1.0-MLX README for the Python API and generation options.
📝 Prompting Guide
Use English prompts. The current open-source models have been validated only with English prompts. Generation quality and prompt adherence in other languages have not been validated. We recommend translating prompts into English before generation.
Write clear, detailed prompts. Specific descriptions of the subject, scene, action sequence, and camera movement can help produce results that better match your intent than short, vague prompts. See the Prompt Guide for structure and examples, and focus on relevant, consistent visual details.
The Prompt Guide describes automatic prompt enhancement in the hosted service. Local examples use your input prompt directly; follow the guide's writing structure and examples when preparing prompts.
📁 Model Files
| Path | Purpose |
|---|---|
transformer/ |
Shared T2V / I2V model |
lora/lite_ti2v_distilled/ |
Lite TI2V Distilled adapter, recipe and defaults |
scheduler_distilled/ |
Lite TI2V Distilled scheduler configuration |
text_encoder_2/, tokenizer_2/ |
Glyph text encoder and tokenizer |
modular_model_index.json defines the required components.
Component provenance
The public model index records component locations and upstream revisions. Text conditioning uses Qwen2.5-VL and a Glyph encoder; image conditioning uses SigLIP, and video decoding uses the Wan VAE. Keep the bundled Glyph encoder and tokenizer with the model.
📜 License
Vimoo model weights and official adapters in this repository are provided under the Apache License 2.0.
- Commercial use, modification, and redistribution are permitted under the terms of the license.
- Redistribution must include a copy of the license, retain applicable copyright and attribution notices, and identify modified files.
- Attribution and third-party component notices are provided in THIRD_PARTY_NOTICES.md. Third-party components remain subject to their own licenses.