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

Tf Icon

Tf Icon is a Python research implementation of TF-ICON, a diffusion-based method for composing user-provided foreground objects into visual contexts, including cross-domain images.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Tf Icon — 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 TF-ICON when you want to explore or build a local, research-oriented workflow that combines a masked foreground with a background, particularly across visual domains.

Who it is for

  • Researchers studying diffusion-based image composition and image inversion.
  • Developers building local image-composition research demos.
  • Users who can prepare image masks and run a Python workflow with GPU resources.

Documented uses

  • Place a foreground object into a selected background and mark the intended placement with a user mask.
  • Explore composition between images from different visual domains.
  • Investigate image inversion as part of a diffusion-based composition workflow.
Before you choose

Requirements & limits

The project is published as a public GitHub repository under the MIT license.

  • The workflow requires four image inputs: background, foreground, a foreground segmentation mask, and a placement mask.
  • The repository recommends 23 GB VRAM and lists 20 GB as a minimum, with needs varying by input.
  • The specified Stable Diffusion 2.1 checkpoint must be downloaded and placed in ./ckpt.
  • Pip installation requires compatible CUDA drivers; CUDA 11.3 is recommended.

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

Prepare TF-ICON for local use

Install the repository environment and place the required Stable Diffusion checkpoint in the expected folder.

Before you start

  • Python environment and the TF-ICON repository.
  • At least 20 GB VRAM is recommended as a minimum; 23 GB is recommended.
  • For pip installation, compatible CUDA drivers are required; CUDA 11.3 is recommended.
  • Stable Diffusion 2.1 checkpoint file named v2-1_512-ema-pruned.ckpt.

Steps

  1. Clone the TF-ICON repository and change into its directory.

  2. Create and activate the environment with Conda using tf_icon_env.yaml, or create a Python virtual environment and install the package with pip install -e .

  3. Download the specified Stable Diffusion checkpoint and put it in the repository’s ./ckpt folder.

Expected outcome

The repository environment is installed and the documented model checkpoint is in place for the subsequent image-composition stage.

Guide limitations

  • The setup instructions do not establish that a particular machine will run the workflow successfully.
  • The recommended VRAM can vary with input samples.
  • The pip options require compatible CUDA drivers.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
17 October 2026
Original website title
Shilin-LU/TF-ICON · Official homepage title · retained in its source language
How we review sourcesSuggest a correction