Read Frog vs KISS Translator: How Two Open-Source AI Translation Extensions Challenge Bloated Paid Plugins

*Source: GitHub repos, official docs, community reviews, and public reporting on the Meituan Tabbit plagiarism incident | Author: Xiaokai | Date: 2026-05-14*

TL;DR: Read Frog and KISS Translator are two open-source browser translation extensions attacking the increasingly bloated, closed-source, privacy-risky Immersive Translate plugin from two directions. Read Frog differentiates with 20+ AI models, context-aware translation, and batched requests saving up to 70% in tokens; KISS Translator wins over technical users with minimalism, broad browser support, offline-first translation, and a custom-rule ecosystem.

• Read Frog repo: https://github.com/mengxi-ream/read-frog
• KISS Translator repo: https://github.com/fishjar/kiss-translator

Why look for alternatives?

Mainstream translation plugins like Immersive Translate suffer from: feature bloat, per-request API token burn, privacy risks (a data leak surfaced in August 2025), freemium/subscriptions, closed source, and page performance degradation from heavy DOM injection. Users increasingly want auditable code, their own API keys, customizable prompts/rules, and offline capability.

Read Frog: AI-driven language learning assistant

| Attribute | Info | |---|---| | Developer | mengxi-ream (ex-ByteDance engineer) | | Stars | ~5.6K (as of 2026-05) | | License | GPLv3 + commercial dual license |

Key features:

• Bilingual / translation-only modes with hot switching, no page refresh
• Context-aware translation: page title + Markdown content summary fed to the AI for more accurate terminology and phrasing
• Selection toolbar: streaming Translate, level-aware Explain, Speak via Edge TTS
• Custom prompts with dynamic tokens ([TARGET_LANG], [INPUT], [TITLE], [SUMMARY])
• Batched requests: merges multiple translation calls into one, claimed up to 70% API cost savings, with exponential-backoff retries and fallback to per-segment requests
• 20+ AI providers via Vercel AI SDK (OpenAI, DeepSeek, Claude, Gemini, Grok, Groq, Mistral, Ollama, plus free Google/Microsoft/DeepLX)
• YouTube bilingual subtitles in the player
• Free Edge TTS: 150+ voices, 80+ languages, sentence-aware chunking with preloading

Tech stack: TypeScript + React + Tailwind CSS + Vite + WebExtension API, IndexedDB (Dexie) storage, fully local operation — users supply their own API keys.

The Meituan Tabbit plagiarism incident

• 2025-12-30: Meituan's Tabbit team forked Read Frog code (at that time no license declared)
• 2026-01-02: The author added GPLv3
• 2026-03-02: Tabbit browser launched its public beta; its translation feature's UI, option order, shortcuts, and even internal filenames (read-frog-xxx.png) were nearly identical to Read Frog
• 2026-03-03: After a public outcry on Xiaohongshu/X, Tabbit admitted the fork, pledged to remove the translation feature and open-source per community norms: https://github.com/Tabbit-Browser/read-frog
• Outcome: The author granted Tabbit a free commercial license (limited to v1.21.3 and earlier)

The episode highlighted the power of GPLv3 copyleft and a growing license-compliance awareness in the Chinese open-source community.

KISS Translator: the minimalist's choice

KISS Translator ("Keep It Simple, Smart") by fishjar is an open-source bilingual translation extension and userscript.

• Browser support: Chrome / Edge / Firefox / Kiwi (Android) / Orion (iOS) / Safari / Thunderbird, plus Tampermonkey/Violentmonkey userscripts
• Services: Google, Microsoft, DeepL/DeepLX, NiuTrans, Baidu, Tencent, Volcengine; AI via OpenAI, Gemini, Claude, Ollama, DeepSeek, OpenRouter, AzureAI, CloudflareAI; Chrome BuiltinAI (offline); custom endpoints
• Scenarios: bilingual page translation, input-box translation, selection translation with multi-service comparison, hover translation, YouTube subtitles with AI segmentation, vocabulary collection
• Offline translation (key differentiator): Browser BuiltinAI (zero config), local LLMs like Ollama (http://localhost:11434), and a 24h local cache
• Advanced: AI context memory, AI terminology dictionaries, batched sending, cross-client sync via KISS-Worker (Cloudflare/Docker) or WebDAV, community rule subscriptions, custom shortcuts

Ecosystem: kiss-translator, kiss-worker, kiss-rules, kiss-proxy, kiss-dictionary.

Head-to-head

| Dimension | Read Frog | KISS Translator | |---|---|---| | Positioning | AI language-learning assistant | Minimalist translation tool | | AI models | 20+ (Vercel AI SDK) | 10+ (incl. custom) | | Batch requests | Yes, claimed 70% token savings | Aggregated sending | | TTS | Edge TTS (free, 150+ voices) | None built-in | | Offline | No | Yes (BuiltinAI / Ollama) | | Browsers | Chrome/Edge/Firefox | 7+ incl. mobile + userscript | | Sync | IndexedDB (Firebase planned) | KISS-Worker / WebDAV | | License | GPLv3 + commercial dual | Open source |

How to choose: language learners → Read Frog; technical docs and privacy-sensitive users → KISS Translator; mobile/multi-device users → KISS Translator; heavy-frequency translators → Read Frog's batching.

Reality check (anti-hype audit)

• The "70% token savings" is a best-case figure; gains are strongest for short, fragmented text and diminish for long articles.
• Neither tool supports PDF translation, EPUB e-books, or AI text polishing — areas where Immersive Translate still leads.
• Open source enables auditing but isn't security itself: users still manage API keys, and browser extensions hold high permissions regardless.

Takeaways

• Token economics: batching short-text requests amortizes system-prompt and HTTP overhead — significant for fragmented web translation.
• Built-in browser AI (Chrome BuiltinAI) is an emerging zero-config, fully local path.
• Decentralization: user-supplied keys, no platform lock-in, community rule ecosystems contrast sharply with closed plugins' lock-in strategies.
• Read Frog may evolve toward a commercialized learning platform; KISS Translator toward translation infrastructure — "the uBlock Origin of translation."

References

1. https://github.com/mengxi-ream/read-frog 2. https://github.com/fishjar/kiss-translator 3. https://blog.mzh.ren/zh/posts/2026/02/meituan-tabbit-plagiarize-read-frog/ 4. https://chrome-stats.com/d/jemckldkclkinpjighnoilpbldbdmmlh