Tooling

Language Learning — Overview & Methodology

How serious learners actually get fluent in 2026 — comprehensible input, sentence mining, SRS, output, and which tools fit which stage.

This is the deep-dive complement to the shorter Language Learning overview. The honest 2026 summary: comprehensible input + spaced repetition + speaking practice beats any single app. Apps are fine for habit-formation; fluency comes from hours of native input, mining the bits you don't know, and producing in the language. This section maps the full free / OSS / self-host landscape plus the few paid tools that genuinely have no free equivalent (Pleco, WaniKani, italki).

Cross-links: Anki SRS · Bilingual Reading · Whisper Transcription · Self-host LLMs · i18n · Voice Dictation.

The four-pillar model

Most modern self-study methodologies (Refold, MIA, Krashen-derived) converge on the same shape:

  1. Comprehensible Input (CI) — listening / reading content just above your level. The Krashen "i+1" hypothesis. Hours matter more than method.
  2. Spaced Repetition (SRS) — Anki + FSRS to lock in the bits CI didn't repeat enough.
  3. Output practice — speaking + writing with feedback (italki, HelloTalk, AI tutors).
  4. Long-term consistency — daily streaks > intense sprints. 1500-2000 hours to functional fluency for a category-1 language (Spanish for English speakers); 3000+ for Japanese / Mandarin.

Stage-appropriate tooling (rough)

  • Pre-A1 / A1 (0-100hrs): Duolingo or LibreLingo for muscle-memory; Pimsleur if you commute; graded readers; FSI courses for free public-domain rigour.
  • A2 / B1 (100-500hrs): start CI (Dreaming Spanish, Easy German, comprehensible podcasts); begin Anki frequency deck; LingQ / Lute for reading.
  • B2 (500-1500hrs): sentence-mine native content with asbplayer + Yomitan / Vocabsieve; italki tutor weekly; native podcasts + YouTube; output journal.
  • C1+ (1500hrs+): native books + media; output-heavy; specialty domains.

OSS-first stack (the serious learner default)

  • SRS: Anki + FSRS (free OSS, AGPL).
  • Mining: asbplayer + Yomitan (JP) or Vocabsieve (universal).
  • Reading: Lute (self-host LingQ alternative).
  • Dictionary: Wiktionary + GoldenDict-NG offline.
  • Listening: Language Reactor (free Chrome ext) on Netflix / YouTube.
  • CI: free YouTube channels per language (Dreaming Spanish, Comprehensible Japanese, Easy German, InnerFrench).
  • Speaking: italki (paid, no free equivalent for live tutors) + HelloTalk (free exchange).
  • Reference: FSI public-domain courses, Tatoeba sentence corpus.

What apps don't do

Duolingo / Babbel / Memrise tap out around A2. They build habits and a passive vocab base, but they don't get you to fluency on their own — confirmed by every honest review of their CEFR calibration studies. Treat them as the gym warm-up, not the workout.

When to pay

  • italki / Preply — there's no free equivalent for cheap live tutors ($5-25/hr).
  • Pleco (Mandarin) — uncontested; the free Lite version is enough for most learners.
  • WaniKani (Japanese kanji) — uniquely effective; some prefer free Anki-only stacks but WaniKani's mnemonic system is the gold standard.
  • AnkiMobile iOS ($25 one-time) — funds the entire free Anki ecosystem.
  • Refold roadmap PDF (~$40) — methodology document; the Discord and YouTube content is free.

Almost everything else has a free or OSS path that's competitive.

The 2024-26 differentiators

  • FSRS replaced SM-2 as Anki's default scheduler (~30% fewer reviews for same retention).
  • asbplayer + Vocabsieve democratised sentence mining (used to require Migaku-level paid tools).
  • Yomitan forked from Yomichan; the canonical Japanese popup dictionary.
  • Refold methodology mainstreamed Krashen's CI hypothesis with concrete tooling.
  • Dreaming Spanish-style CI YouTube exploded across languages — 1000+ hours of free beginner-graded video for Spanish, French, German, Thai, Mandarin.
  • AI tutors (Talkpal, Speak) are decent for low-stakes practice but don't replace humans.
  • FSI public-domain courses rediscovered — 50+ languages, free, surprisingly rigorous.

Pick this if…

On this page