OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

Zi2Zi

Zi2Zi is an open-source Python project that adapts pix2pix with a conditional GAN to learn East Asian character styles from paired font images and generate glyphs in a target style. It includes scripts for preparing character-image data, training a model, running inference and interpolating between font labels.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Zi2Zi — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Zi2Zi when you want to experiment with learning character styles from font-image pairs and can work with its documented Python-based training and inference workflow.

Who it is for

  • Researchers exploring conditional GANs and character-style transfer
  • Developers prototyping Chinese or other East Asian glyph-generation workflows
  • Typeface designers experimenting with learned font-style variants

Documented uses

  • Prepare paired character images from source and target fonts for training.
  • Train a model conditioned on font-category labels.
  • Generate character glyphs from a trained checkpoint.
  • Explore transitions between selected font styles through interpolation.
Before you choose

Requirements & limits

The code is available in a public GitHub repository under the Apache-2.0 license.

  • The documented workflow requires font files, preprocessing and packaged image data rather than a ready-to-use hosted interface.
  • The README specifies a legacy stack including Python 2.7 and TensorFlow 1.0.1 or later, plus CUDA and cuDNN.
  • The project offers four default character sets: simplified Chinese, traditional Chinese, Japanese and Korean.
  • An optional label-shuffling method is intended for further tuning after discriminator loss plateaus near zero.

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 paired font data for training

Convert source and target font files into packaged character-image data for a Zi2Zi experiment.

Before you start

  • Python 2.7, CUDA, cuDNN and TensorFlow 1.0.1 or later
  • Pillow, NumPy 1.12.1 or later, SciPy 0.18.1 or later and imageio
  • Source and target font files

Steps

  1. Run font2img.py with source and target font files, a character set, sample count, output directory and font label.

  2. Run package.py on the image directories and choose a save directory and split ratio.

  3. Create an experiment/data directory and place the resulting train.obj and val.obj files there.

Expected outcome

Training and validation binary data are arranged in the experiment directory expected by the documented training workflow.

Guide limitations

  • The README lists legacy Python and TensorFlow requirements.
  • The guide covers data preparation and layout, not model training or inference.
Documentation 1 ↗
Source-based practical guide

Generate glyphs with a trained model

Run inference on character data using an existing Zi2Zi checkpoint and selected font labels.

Before you start

  • A trained model checkpoint
  • A binary character-data object
  • The project's documented Python, CUDA, cuDNN, TensorFlow, Pillow, NumPy, SciPy and imageio requirements

Steps

  1. Run infer.py with the checkpoint directory, batch size, source object, font embedding label or labels, and output directory.

  2. For style interpolation, set the interpolation option, provide the font labels and number of steps, and specify frame and GIF output paths.

Expected outcome

The inference workflow writes generated results to the chosen output location; interpolation can produce frames and a GIF.

Guide limitations

  • This guide assumes data preparation and model training have already been completed.
Documentation 1 ↗
COMPARE OPTIONS

Compare options for Model inference and serving

These projects document a common task and share a project role. This is a source-based comparison, not an execution benchmark or a guarantee of interchangeability.

ProjectDocumented purposePlatformAccess and requirements
Zi2ZiPrepare paired character images from source and target fonts for training.LinuxThe code is available in a public GitHub repository under the Apache-2.0 license.
Pulse Of MotionEstimate a video's physical frame rate from its visual dynamics.LinuxThe repository is public. Its README describes a released checkpoint that downloads from Hugging Face on first inference, with a manual download option. The strongest internal model is not fully released; the README directs interested users to contact the project about access. No license information is provided in the repository evidence.
Neural Image AssessmentTrain the model using AVA images and rating annotations.Cloud, Linux, macOS, Windows
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
16 October 2026
Original website title
kaonashi-tyc/zi2zi · Official homepage title · retained in its source language
How we review sourcesSuggest a correction