thomaswantstobeaskeleton/BallonsTranslator-Pro
thomaswantstobeaskeleton/BallonsTranslator-Pro has 100 GitHub stars and 10 forks. GitHot measured 0 stars in 24 hours. Last checked 2026-11-15 16:18 UTC. Last push 2026-07-31.
Measured star growth
- in 24h
- 0
- in 7 days
- Update delayed
- in 15 days
- Update delayed
- in 1 month
- Update delayed
Net star changes over overlapping periods, not daily totals. Missing measurements are left blank. Stars measure attention, not software quality.
About
基于 BallonsTranslator 的漫画/网漫/韩漫/国漫计算机辅助翻译工具,扩展了 OCR、文本检测、|修复、工作流程、字体和导出选项。Computer-aided manga/comic/manhwa/manhua translation tool based on BallonsTranslator, with expanded OCR, detection, inpainting, workflow, font, and export options.
Machine LearningREADME.md
Original[!IMPORTANT] If you're sharing the translated result publicly and no experienced human translator participated in a thorough translation or proofread, please mark it as machine translation somewhere clear to see.
BallonsTranslator-Pro
AI-powered manga & comic translation toolkit — forked from dmMaze/BallonsTranslator with 50+ extra modules and production-grade workflow tools.
简体中文 | English | pt-BR | Русский | 日本語 | Indonesia | Tiếng Việt | 한국어 | Español | Français
What is BallonsTranslator-Pro?
BallonsTranslator-Pro is an advanced fork of dmMaze/BallonsTranslator built for serious manga/comic scanlation workflows. It keeps everything great from the original and adds:
- 90+ AI modules — 20+ text detectors, 30+ OCR engines, 25+ translators, 15+ inpainters
- Production lettering tools — auto-layout, overflow checks, shape-aware safe areas, density-aware font scaling
- Realtime screen translation — always-on-top overlay for live manga reader translation
- Translation Assist dock — per-block TM/glossary/candidates, provider comparison, QA warnings
- Batch queue — multi-folder processing with pause/resume/cancel
- Automation API — local REST API + MCP-style command surface for headless workflows
- Docker / server mode — run without GUI for CI or remote workflows
Quick Start
| OS | One-liner |
|---|---|
Double-click launcher.bat or launch_win.bat |
|
python launch.py |
|
python launch.py |
# Or clone and run
git clone https://github.com/thomaswantstobeaskeleton/BallonsTranslator-Pro
cd BallonsTranslator-Pro
python launch.py
GPU help (AMD RX 9070/9060/7900, NVIDIA, Apple Silicon): docs/GPU_ACCELERATION.md
Feature Highlights
One-Click Translation Pipeline

Mix-and-match 90+ modules across the full pipeline. No need to settle for one-size-fits-all defaults.
Realtime Screen Translator
Always-on-top, click-through overlay. Select any screen region or follow a window. Privacy-first: nothing is persisted or logged by default.
Translation Assist
Per-block candidate suggestions, TM/glossary lookup, side-by-side provider comparison, and block-level QA warnings.
Production Lettering & Auto-Layout

Shape-aware safe areas, balanced line breaks, density-aware font scaling, and overflow safety checks.
Batch Queue
Process multiple folders in a single queue with pause, resume, and cancel.
Module Catalog
Text Detectors (20+) — click to expand
| Module | Type | Best For |
|---|---|---|
ctd (ComicTextDetector) |
Built-in | Default; Japanese/English manga |
ysgyolo |
Pro | Filters onomatopoeia; lh5426 model |
animetext_yolo |
Pro | Complex anime scenes (AnimeText dataset) |
mangalens_bubble_segmentation |
Pro | Bubble segmentation |
paddle_det / paddle_det_v5 |
Pro | Chinese/English/Japanese |
pp_doclayout_v3 |
Pro | Complex layouts, curved pages, mixed columns |
easyocr_det |
Pro | Multilingual scene text |
craft_det |
Pro | Curved/oriented text |
mmocr_det |
Pro | Document/comic text (polygon) |
rapidocr_det |
Pro | Lightweight ONNX detection |
dptext_detr |
Pro | Dynamic point queries |
hf_object_det |
Pro | Bubble/text object detection |
hunyuan_ocr_det |
Pro | Full-image spotting |
magi_det |
Pro | Manga panels + reading order (CVPR 2024) |
surya_det |
Pro | 90+ languages, line-level |
sam_text_det / sam3_refiner |
Pro | SAM prompt-based segmentation |
swintextspotter_v2 |
Pro | End-to-end text spotting |
textmamba_det |
Pro | Curved text (Mamba SSM) |
stariver_ocr |
Pro | Stariver cloud OCR |
OCR Engines (30+) — click to expand
| Module | Type | Best For |
|---|---|---|
mit_32px / mit_48px / mit_48px_ctc |
Built-in | Manga-image-translator models |
manga_ocr |
Built-in | Japanese manga (kha-white) |
paddle_ocr / paddle_vl |
Pro | Chinese/Japanese/English |
easyocr_ocr |
Pro | Multilingual |
rapid_ocr |
Pro | Lightweight ONNX |
florence2_ocr |
Pro | Microsoft vision model |
got_ocr2 |
Pro | Unified plain/scene/formatted |
hunyuan_ocr |
Pro | 100+ languages, spotting |
glm_ocr |
Pro | 0.9B document OCR |
internvl2_ocr / internvl3_ocr |
Pro | Document/chart/OCR |
lighton_ocr |
Pro | 1B parameter OCR |
chandra_ocr |
Pro | 9B document OCR (layout, tables, math) |
deepseek_ocr |
Pro | Heavyweight document OCR |
docowl2_ocr |
Pro | OCR-free document understanding |
bing_ocr |
Pro | Bing image OCR API |
google_vision |
Pro | Google Cloud Vision |
google_lens_exp |
Pro | Experimental Google Lens API |
macos_ocr |
Pro | Apple Vision (macOS native) |
callisto_ocr / qwen2vl_ocr |
Pro | VLM-based 2B OCR |
vlm_ocr (generic HF) |
Pro | Any HF VLM (Qwen, InternVL, OlmOCR) |
donut |
Pro | OCR-free document understanding |
llm_ocr |
Pro | LLM API OCR |
lens_proto |
Pro | Google Lens Protobuf |
Translators (25+) — click to expand
| Module | Type | Best For |
|---|---|---|
google |
Built-in | Free, fast |
deepl / deeplx / deeplx_api |
Pro | High-quality NMT |
sugoi |
Pro | Japanese → English (offline) |
sakura |
Pro | Japanese → Chinese (Sakura-13B) |
chatgpt / chatgpt_exp / openai |
Pro | GPT-4 / GPT-3.5 |
gemini_neverliie / mistral_neverliie |
Pro | Gemini / Mistral (neverliie SDK) |
cohere_command_r |
Pro | Cohere Command R+ |
qwen_mt |
Pro | Alibaba Qwen translation |
hy_mt_1_5_7b |
Pro | Tencent Hunyuan MT |
hunyuan_mt_chimera_7b |
Pro | Ensemble refiner (multi-source → Chimera) |
chimera |
Pro | Multi-source ensemble |
ensemble |
Pro | 3 translators + LLM judge |
chain |
Pro | Chained translation (e.g. JP → EN → CN) |
mbart50 |
Pro | 50-language Meta NMT |
nllb200 |
Pro | 200-language Meta NMT |
opus_mt |
Pro | Helsinki NLP per-pair models |
t5_mt |
Pro | Prompt-based T5 translation |
m2m100 / m2m100_hf |
Pro | Meta many-to-many |
manual |
Pro | JSON prompt workflow |
llm_api_translator |
Pro | Generic LLM API |
eztrans |
Pro | Korean game translation |
text-generation-webui |
Pro | Local TGW backend |
translatorspack |
Pro | Aggregator (google, bing, baidu, etc.) |
caiyun / baidu / papago |
Pro | Chinese cloud APIs |
Inpainters (15+) — click to expand
| Module | Type | Best For |
|---|---|---|
aot |
Built-in | Manga-image-translator default |
patchmatch |
Built-in | Non-deep learning (PS-like) |
lama_mpe |
Built-in | Fine-tuned LaMa |
cuhk_manga_inpaint |
Pro | CUHK Seamless Manga (SIGGRAPH 2021) |
lama_onnx / lama_manga_onnx |
Pro | ONNX LaMa (general / manga) |
simple_lama |
Pro | pip-installable LaMa |
mat |
Pro | Mask-Aware Transformer (CVPR 2022) |
opencv-tela / opencv-classic |
Pro | OpenCV inpainting |
diffusers_sd_inpaint / diffusers_sd2_inpaint / diffusers_sdxl_inpaint |
Pro | Stable Diffusion family |
dreamshaper_inpaint |
Pro | DreamShaper 8 |
kandinsky_inpaint |
Pro | Kandinsky 2.1 |
fluently_v4_inpaint |
Pro | Anime/comic tuned |
repaint |
Pro | DDPM-based |
flux_fill |
Pro | FLUX.1 Fill (12B, high quality) |
qwen_image_edit |
Pro | Qwen-Image-Edit semantic fill |
Pro vs Original — At a Glance
| Capability | Original | Pro |
|---|---|---|
| Text Detectors | 3 | 20+ |
| OCR Engines | 5 | 30+ |
| Translators | 10 | 25+ |
| Inpainters | 3 | 15+ |
| Realtime Screen Overlay | No | Yes |
| Translation Assist Dock | No | Yes |
| Batch Queue (multi-folder) | No | Yes |
| Layout Review & Auto-Layout | Basic | Advanced |
| Automation API / Headless | No | Yes |
| Docker / Server Mode | No | Yes |
| In-App Google Fonts Installer | No | Yes |
| Environment Doctor | No | Yes |
| Model Manager | No | Yes |
| PSD Export | No | Yes |
| InDesign LPTXT Workflow | No | Yes |
Basic Workflow
- Open pages —
File → Open Folder / Open Images - Select modules — pick your detector + OCR + translator + inpainter
- Run — one-click full pipeline
- Review & edit — rich text editor with undo/redo, multi-select, style presets
- Export —
Tools → Export all pages(PNG, CBZ, PSD, InDesign LPTXT)
System Requirements
- Python 3.10.2+ (avoid 3.10.0/3.10.1 — PyInstaller crash bug)
- Internet for first-time setup and model downloads
- Disk ~10-50 GB depending on selected models
- GPU optional (CUDA, ROCm, Apple Silicon, or CPU)
python --version
Documentation
| Start Here | Quality & Lettering | Automation & Plans |
|---|---|---|
| TROUBLESHOOTING | QUALITY_RANKINGS | LOCAL_AUTOMATION_API |
| GPU_ACCELERATION | MODELS_REFERENCE | REALTIME_TRANSLATION_MODE_PLAN |
| DOCS_HIGHLIGHTS | TRANSLATION_CONTEXT_AND_GLOSSARY | TRANSLATION_ASSIST_PLAN |
| INDESIGN_LPTXT_WORKFLOW | FEATURE_PARITY_MATRIX |
- Full change history: docs/CHANGELOG.md
Manga / Comic Raw Downloader
BallonsTranslator-Pro includes a registry-backed manga source downloader with built-in providers (MangaDex, Comick, GOMANGA, MangaForFree, ToonGod, Mangakakalot, NaruRaw, and more). New providers can be added via utils/manga_sources/provider_base.py without rebuilding the UI.
docs/RAW_SOURCE_PROVIDER_EXPANSION_PLAN.md— roadmapdocs/MANGA_SOURCE_PLUGIN_API.md— provider interface
python scripts/test_manga_sources.py
pytest -q tests/test_manga_provider_base.py
License
GPL-3.0 — Forked from dmMaze/BallonsTranslator.