Does Homie work in the EU?

Available in the EU. Reportedly released as open weights under Apache 2.0, which has no regional restrictions, so it can be self-hosted in the EU. No official EU-specific statement was found, and the weights' location was not verified. Last verified 14 Nov 2026.

How much does Homie cost?

Open source. Reported as Apache 2.0 with downloadable weights (about 37 GB); not verified on an official repository. Licence: Apache-2.0.

Where was Homie shown?

In 1 creator video by AI Search. Most recently by AI Search on 26 Jul 2026, at 6:28 in "Claude Opus 5, GPT 6 hack, Flux 3, new Gemini, quantum breakthrough, new Qwen: AI NEWS".

Homie

Website

Reference-to-video model from HKUST for personalizing videos with multiple people and objects.

Available in EUVideo GenerationModel

Updated 14 Nov 2026EU status checked 14 Nov 2026Last shown 26 Jul 2026

Homie at a glance

What is Homie?: Homie is a model. Reference-to-video model from HKUST for personalizing videos with multiple people and objects.
Does Homie work in the EU?: Available in the EU. Reportedly released as open weights under Apache 2.0, which has no regional restrictions, so it can be self-hosted in the EU. No official EU-specific statement was found, and the weights' location was not verified. Last verified 14 Nov 2026.
How much does Homie cost?: Open source. Reported as Apache 2.0 with downloadable weights (about 37 GB); not verified on an official repository. Licence: Apache-2.0.
Where was Homie shown?: In 1 creator video by AI Search. Most recently by AI Search on 26 Jul 2026, at 6:28 in “Claude Opus 5, GPT 6 hack, Flux 3, new Gemini, quantum breakthrough, new Qwen: AI NEWS”.

Sources: the vendor's own pages (linked above) and creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.

About

HOMIE (Human-object Centric Video Personalization) is a research project from the Hong Kong University of Science and Technology. It generates videos from reference images of people and objects, aiming to keep the identity of several subjects consistent within one clip.

According to the demo page, it covers inter-subject personalization of human-object interaction, logo personalization using multimodal LLM reasoning, intra-subject personalization with multi-view consistency, and high text fidelity using OCR maps. A YouTube review describes it as built on Wan 2.1 and Phantom, released under Apache 2.0, with roughly 37 GB of weights.

The provided material is a demo page only; no GitHub, Hugging Face or paper links were found in the fetched text, so release date and download location are unconfirmed. It is aimed at researchers and developers able to run large video models locally.

Key features

Reference-to-video generation with multiple people and objects
Multi-view reference support for consistency
OCR maps for text and logo fidelity
MLLM reasoning for logo personalization
Built on Wan 2.1 and Phantom

1 moment

As seen on YouTube

Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.

AI Search26 Jul 2026

Live demo

Claude Opus 5, GPT 6 hack, Flux 3, new Gemini, quantum breakthrough, new Qwen: AI NEWS

Reference-to-video generator preserving consistency of multiple people and objects, supports multi-view references and OCR maps for text. Built on Wan 2.1 and Phantom, Apache 2 license, 37 GB.

6:28Open on YouTube