language-learning/README.md

74 lines
4.3 KiB
Markdown

# Language Learning Research
Collection of open-source language learning tools and resources for evaluation.
## Cloning
This repo uses git submodules. To get everything in one shot:
```bash
git clone --recurse-submodules https://github.com/YOU/language-learning.git
```
If you already cloned without `--recurse-submodules`:
```bash
git submodule update --init --recursive
```
## What's In Here
### repos/
| Repo | What It Does |
| -------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------- |
| [companion](https://github.com/shakedzy/companion) | Private AI language tutor — Whisper + ChatGPT + Google TTS. Read, write, talk, listen in native and target language. |
| [discute](https://github.com/5uru/Discute) | Speaking practice app — designed specifically for conversational practice in a new language. |
| [ai-english-tutor](https://github.com/jasonkang14/ai-english-tutor) | Interactive AI English tutor using OpenAI APIs. |
| [linguacafe](https://github.com/simjanos-dev/LinguaCafe) | Self-hosted web app for reading-based vocabulary acquisition. 27+ languages. Docker-based. |
| [librelingo](https://github.com/kantord/LibreLingo) | Community-driven open language learning platform. Course editor + web app. |
| [libretranslate](https://github.com/LibreTranslate/LibreTranslate) | Self-hosted machine translation API. No Google/Azure dependency. |
| [speechbrain](https://github.com/speechbrain/speechbrain) | Conversational AI toolkit — speech recognition, speaker verification, TTS. Building block for custom tools. |
| [awesome-language-learning](https://github.com/Vuizur/awesome-language-learning) | Curated list of open-source language learning projects and resources. |
### Speaking-Focused (most relevant to mic-based learning)
- **companion** — closest to a full "talk to practice" solution out of the box
- **discute** — purpose-built for speaking practice
- **ai-english-tutor** — English-specific but demonstrates the tutor pattern
### Infrastructure / Building Blocks
- **speechbrain** — STT/TTS toolkit if building something custom
- **libretranslate** — self-hosted translation backend
- **linguacafe** — reading/vocab complement to speaking practice
### Reference
- **awesome-language-learning** — index of more projects worth exploring
## Commercial Apps Worth Knowing About
These can't be cloned but are worth evaluating alongside the open-source options:
- [ChatGPT Voice Mode](https://chatgpt.com) — free-form conversation in any language (needs Plus sub)
- [Langua](https://langua.io) (~$8-12/mo) — native-speaker-cloned voices, 23 languages, real-time feedback
- [Talkio AI](https://www.talkio.ai) (~$10-15/mo) — 40+ languages, 134 dialects, pronunciation feedback
- [Speak](https://www.speak.com) (~$15-20/mo) — structured curriculum, good for beginners
- [ELSA Speak](https://elsaspeak.com) (~$6-10/mo) — phoneme-level pronunciation analysis (English only)
## The DIY Stack
If none of these fit, the common architecture for a custom speaking tutor:
1. **STT**: [Whisper](https://github.com/openai/whisper) (local or API)
2. **LLM**: [Claude API](https://docs.anthropic.com/en/docs) or [GPT-4o](https://platform.openai.com/docs) with a tutor system prompt
3. **TTS**: [OpenAI TTS API](https://platform.openai.com/docs/guides/text-to-speech), [Coqui TTS](https://github.com/coqui-ai/TTS) (open source), or speechbrain
## What Linguists Recommend (2025-2026)
- Massive input first (listening + reading), then speaking practice
- AI conversation daily for low-pressure reps
- Human tutor 1-2x/week for nuance
- Spaced repetition for vocab ([Anki](https://apps.ankiweb.net), etc.)
- Consistency over intensity