LivePortrait

Overview

LivePortrait is a free, open-source system for animating faces and editing portrait video. It suits creators and developers who want control over motion inputs and can work with a local AI setup. Its flexibility is the draw; installation requirements and licensing conditions for included detection models are important trade-offs.

It animates a still portrait using motion from video, audio, text, or generation, and supports video editing with a source video and a driving video. Face animation is supported, but depth animation is not; MP4 is the stated export format. A Hugging Face Spaces demo provides an online way to try the project, while local use relies on inference code, pretrained weights, and a Gradio interface.

Key features

• Multiple motion inputs: Video, audio, text, or generation can drive a portrait. That range suits users who want to shape facial movement in different ways, rather than choose only from fixed effects.
• Retargeting and stitching: Eye- and lip-retargeting modules and stitching controls offer finer adjustment of facial animation. They make LivePortrait more appealing for users seeking control, though they do not remove the possibility of visual artifacts.
• Portrait video editing: Pairing a source video with a driving video extends the project beyond animating a single still image.
• Animal animation: Fine-tuned models can animate cats, dogs, and pandas using human driving motion. This is a distinctive creative option, but animal mode is limited to Linux and Windows with an NVIDIA GPU; it is unsupported on macOS because X-Pose does not support that platform.
• Reported speed: The project reports generation at 12.8 ms on an RTX 4090 GPU with PyTorch. That figure applies to the named hardware and framework, so it should not be treated as a general speed guarantee.
• Privacy option: Motion templates can speed inference and protect privacy, giving users an alternative to using a driving video directly.

Pricing

LivePortrait is free and open source under the MIT License, with no paid plans described. Local use does require git, conda, FFmpeg, Python 3.10, and downloaded pretrained weights, so the absence of a software charge does not make setup effortless.

Commercial use needs particular care: InsightFace detection models included in the project are for non-commercial research. Commercial users must remove and replace those models to comply with their license conditions; the MIT license for the implementation does not remove that requirement.

Platforms

Human-mode use is documented for Linux, Windows, and macOS with Apple Silicon. Intel macOS is untested in the stated instructions, so users on that hardware should not assume a supported local workflow. The online demo offers a route that avoids local installation, while the project also lists community integrations including FaceFusion, ComfyUI nodes, a Stable Diffusion WebUI extension, Replicate, FAL, Poe, and Hugging Face Spaces.

Animal mode has narrower hardware support than human mode: it is tested only on Linux and Windows with an NVIDIA GPU and is unsupported on macOS.

Who it's for

LivePortrait is a strong fit for technically comfortable creators and developers who want a free, locally runnable portrait-animation project with multiple motion inputs and facial controls. It can also suit users exploring animal animation, provided they have the required Linux or Windows system and NVIDIA GPU. It is a weaker fit for anyone who needs a simple, fully managed application, dependable Intel Mac support, or animal animation on macOS.

Pros and cons

• Pro: Video, audio, text, and generated motion inputs give users several ways to drive a portrait.
• Pro: Stitching and eye- and lip-retargeting controls provide more adjustment than a basic animation pass.
• Pro: Free MIT-licensed implementation and local inference code make it accessible to users who can manage the setup.
• Con: Local installation depends on several tools, a specific Python version, and downloaded model weights.
• Con: Commercial users must replace the included InsightFace detection models.
• Con: Animal mode requires Linux or Windows and an NVIDIA GPU, excluding macOS workflows.
• Con: The maintainers warn of deepfake misuse and possible visual artifacts, and disclaim legal responsibility for generated results.

Alternatives

Photo Animation Software and AI Photo Animation Tools are broader directories for comparing options in those categories.

• Plotaverse is worth considering for users who want a freemium option across Android, iOS, macOS, and web, rather than a local open-source project.
• Deep Nostalgia suits users looking for several free animations and access on Android, iOS, or web.
• Parallax Studio is another free, open-source choice for Linux, macOS, or Windows users comparing locally oriented software.
• Zoetropic is an option for Android and iOS users seeking basic motion and camera effects in a freemium app.
• Hedra may suit web or API users who prefer a monthly plan with a stated credit allowance and commercial use.
• Higgsfield offers web and API access with paid monthly plans measured in credits, for users who prefer that model.
• Free.ai Text to Video is an API option for users focused on text-to-video generation, with daily free tokens and a limit of one clip every two days on its free plan.
• Immersity AI is an API alternative for users seeking free, unlimited conversions up to 720p who can accept watermarks and non-commercial use.

Verdict

Choose LivePortrait if you want free, controllable portrait animation and are prepared to manage local AI tooling; its varied motion inputs and facial controls are the clearest reasons to choose it. Look elsewhere if you need a simpler hosted workflow, broad Mac support for animal animation, or commercial use without replacing the included detection models.