Hotshot-XL

Paid Self-hostedWeb

What Hotshot-XL is

Hotshot-XL is a free, open-source model for generating GIFs from text prompts, designed to work with Stable Diffusion XL (SDXL). It can use existing or newly fine-tuned SDXL models, including user-provided LoRAs for personalized subjects. SDXL ControlNet can guide composition and layout, with one ControlNet model supported at a time. The model is trained for one-second GIFs at 8 frames per second. It supports several trained aspect ratios from 320×768 through 768×320; fine-tuning is recommended for reliable results beyond that range. Changing the frame rate or duration is experimental and can produce unstable or jittery results. The project provides instructions for installing dependencies, downloading weights, and running inference locally, plus a fine-tuning script for additional text and video pairs. Its model card notes limits with photorealism, legible text, complex composition, and rendering faces and people. The model has 3B parameters. The GitHub repository lists Apache-2.0, while the model page lists Open RAIL++-M.

Who it suits

It suits people seeking a self-hosted or web-based way to turn prompts into short GIFs using SDXL. Its one-second default and experimental variable duration may not suit projects needing reliably longer animations.

What it does well

Uses existing or fine-tuned SDXL models.
Accepts personalized SDXL LoRAs.
Supports SDXL ControlNet guidance.
Includes local inference instructions and a fine-tuning script.

What to know first

Default output is one second at 8 FPS.
Variable duration or frame rate is experimental.
Only one ControlNet model can be active at once.
Model card notes limits with text, composition, and people.

Limits

The trained output is a one-second GIF at 8 FPS, and only one ControlNet model is supported at once. Changing duration or frame rate is experimental and may result in unstable or jittery output; reliable generation outside the trained aspect ratios may require fine-tuning.

Getting started

Open the Hotshot-XL GitHub project page.
Follow its instructions to install dependencies and download model weights.
Run inference locally using an SDXL model.
Use an SDXL LoRA for personalized subjects or one SDXL ControlNet model for composition guidance.

Verdict

Hotshot-XL offers prompt-to-GIF generation tied to SDXL, with LoRA personalization and ControlNet guidance. Its trained output is short, and changing duration or frame rate may lead to unstable results.

Questions

Is Hotshot-XL free?

Yes. It is listed as free and open source.

What platforms does it support?

The listed platforms are self-hosted and web. The project provides instructions for running inference locally.

What does it generate?

It generates GIFs from text prompts and is designed to work alongside Stable Diffusion XL.

Can I personalize the output?

Yes. It can use user-provided SDXL LoRAs, and the project recommends them for personalized subjects.

What licenses are listed?

The GitHub repository lists Apache-2.0, while the Hugging Face model page identifies Open RAIL++-M.

Facts from Hotshot-XL's own pages

Aspect ratios: It supports several trained aspect ratios, from 320×768 through 768×320, and recommends fine-tuning for reliable generation outside that range. · github.com, 9 Oct 2026
ControlNet: It supports SDXL ControlNet for guiding GIF composition and layout, with one ControlNet model supported at a time. · github.com, 9 Oct 2026
ControlNet limit: The README says Hotshot-XL supports one ControlNet model at a time. · github.com, 9 Oct 2026
Default output: The model was trained to generate one-second GIFs at 8 FPS. · github.com, 9 Oct 2026
Fine-tuning: The repository includes a fine-tuning script for additional text and video pairs and recommends SDXL LoRAs for personalized subjects instead of fine-tuning Hotshot-XL. · github.com, 9 Oct 2026
GIF format: The model was trained to generate one-second GIFs at 8 FPS. · github.com, 9 Oct 2026
License: The repository lists the Apache-2.0 license. · github.com, 9 Oct 2026
Local use: The project provides instructions to install dependencies, download model weights, and run inference locally. · github.com, 9 Oct 2026
Maximum GIF length: 1 · github.com, 22 Sept 2026
Model limitations: The model card says the model does not achieve perfect photorealism, cannot render legible text, struggles with compositional prompts, and may generate faces and people improperly. · huggingface.co, 9 Oct 2026
Model size: The Hugging Face model page lists the model as having 3B parameters. · huggingface.co, 9 Oct 2026
Personalization: It can generate GIFs with fine-tuned SDXL models and load user-provided SDXL LoRAs. · github.com, 9 Oct 2026
Prompt to GIF: Yes · github.com, 22 Sept 2026
Purpose: Hotshot-XL is an AI text-to-GIF model designed to work alongside Stable Diffusion XL. · github.com, 9 Oct 2026
SDXL models: It can generate GIFs with existing or newly fine-tuned SDXL models, including personalized SDXL LoRAs. · github.com, 9 Oct 2026
Support and contributions: The README invites users to raise issues or pull requests, and links to a Discord community. · github.com, 9 Oct 2026
GIF templates: No · github.com, 22 Sept 2026
Training integration: The fine-tuning instructions describe sending validation GIFs to Weights & Biases. · github.com, 9 Oct 2026
Variable duration: Changing frame rate or duration is experimental, and the project warns that results may be unstable or jittery because training used one-second videos at 8 FPS. · github.com, 9 Oct 2026
Variable length: Changing frame rate or duration is experimental, and the README warns results may be unstable or jittery because training used one-second videos at 8 FPS. · github.com, 9 Oct 2026
Weights & Biases: The README describes sending validation GIFs to Weights & Biases as an optional fine-tuning workflow. · github.com, 9 Oct 2026
What it does: Hotshot-XL is an AI text-to-GIF model trained to work alongside Stable Diffusion XL. · github.com, 9 Oct 2026

Where Hotshot-XL stands on Freedom251

GIF makers#51 of 54
AI GIF Makers#23 of 23