# kaonashi-tyc/zi2zi

Learning Chinese Character style with conditional GAN

Repository: https://github.com/kaonashi-tyc/zi2zi
Canonical: https://ross.abutalabs.com/products/zi2zi
Homepage: https://kaonashi-tyc.github.io/2017/04/06/zi2zi.html
Language: Python
License: Apache-2.0
License Family: permissive
Topics: deeplearning, generative-adversarial-networks, pix2pix, tensorflow, machine-learning, chinese-characters, style-transfer, deep-learning, deep-neural-networks
Last push: 2019-08-09T04:02:31+00:00

## Health v2 (maintenance only)
Score: 32/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 3484, "days_push": 2581, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2745, forks 475 (observed 2026-08-28T04:07:16.761228+00:00)

## What it is
zi2zi is a TensorFlow implementation of a conditional GAN (based on pix2pix) for learning and transferring Chinese and East Asian character typefaces. It generates new font styles from example fonts using category embeddings, category and constant losses, and optional label shuffling.

## Use cases
- generate a new Chinese calligraphy font from a few sample characters
- convert text rendered in one font style to another style
- create brush writing or cursive script typefaces automatically
- generate Korean hangul typefaces from Chinese font training
- interpolate between font styles for animation effects

## When to choose
- you need to synthesize East Asian typefaces in new styles from limited samples
- you want a research baseline or reference implementation of conditional GAN font style transfer
- you are working with TensorFlow 1.x and have CUDA GPUs available

## When to avoid
- you need a maintained library or support for modern TensorFlow versions
- you want to generate Latin or non-East-Asian scripts without adaptation
- you lack GPU resources or need production-ready font generation

## Facets
- artifact type: library
- maturity: abandoned
- function: machine-learning, deep-learning, image-processing
- domain: machine-learning, deep-learning, graphics
- platform: python
- tags: gan, pix2pix, style-transfer, chinese-characters, calligraphy, font-generation, tensorflow, natural-language-processing, gpu, linux

## Member repositories
- kaonashi-tyc/zi2zi (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:16.761228+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T08:19:04.580832+00:00, confidence not recorded.
  - readme: https://github.com/kaonashi-tyc/zi2zi (fetched 2026-08-28T04:07:16.761228+00:00, sha 80e096039b7d)
  - homepage: https://kaonashi-tyc.github.io/2017/04/06/zi2zi.html (fetched 2026-08-29T09:57:07.167564+00:00, sha fc7a8cb9d2de)
- Data as of 2026-08-30T08:39:29.467469+00:00.