TutorFlow is built to handle the four core language skills (listening, speaking, reading, and writing) inside a single authoring flow. Rather than stitching together flashcard decks, audio players, recording tools, and grading rubrics across separate apps, you can generate a full skill-balanced course from one prompt and publish it to your classroom.
This makes TutorFlow practical for foreign language instructors, K-12 and university language departments, corporate language training programs, and independent tutors who need to move faster than traditional textbook workflows allow.
What makes language courses different in TutorFlow
A language course is rarely served well by generic lesson types. Listening drills need audio playback, speaking drills need microphone recording, reading drills need passage comprehension, and writing drills need nuanced grading that accepts synonyms and stylistic variation.
TutorFlow ships five language-specialized lesson types that each target a specific skill combination, plus flashcards for vocabulary. Every type works across any language the platform supports, with first-class handling for CJK (Chinese, Japanese, Korean) scripts.
| Primary skill | Lesson types you can use |
|---|---|
| Listening | Dictation, Shadowing |
| Speaking | Shadowing, Roleplay |
| Reading | Reading Comprehension, Translation |
| Writing | Dictation, Translation |
| Vocabulary | Flashcards (included in the language category) |
A well-designed language chapter usually mixes three or four of these so learners practice more than one skill in a single sitting.
The five language practice lesson types
Dictation
Listening plus writing. The platform plays a target-language sentence through text-to-speech and the learner types exactly what they hear. The grader compares the transcript to the reference sentence and scores accuracy, with optional lenient mode that forgives minor punctuation and typo differences.
Dictation is the fastest way to train listening accuracy alongside spelling and script recognition, and it works at any CEFR level. For CJK languages you can enable pinyin or furigana guides next to the reference so learners can check their reading as well as their writing.
Shadowing
Listening plus pronunciation practice. A native-sounding audio line plays, the learner taps record and repeats it, and the recording is transcribed with Whisper and compared to the target line for a similarity score. The language is pinned explicitly, so a learner who accidentally slips into their first language still gets scored against the target language.
Shadowing is ideal for pronunciation, rhythm, and intonation training. It is the closest TutorFlow comes to a real tutor listening to each learner, and it scales to entire cohorts without any manual grading.
Roleplay
Multi-turn conversation with an AI character against a scenario and a set of success criteria. A typical roleplay looks like ordering at a Parisian cafe, checking into a hotel in Tokyo, or running a job interview in English. The learner sees the scene, speaks or types each turn, and the AI character responds in character until the scenario concludes. At the end, an AI evaluator rates the conversation on task completion, language quality, and situational appropriateness against the criteria you defined.
Roleplay targets speaking, pragmatics, and situational fluency, and it is the closest simulation of real-world language use the platform offers. It works best at B1 and above, but the difficulty is adjustable through the scenario prompt.
Reading Comprehension
A short reading passage followed by a mix of multiple-choice questions and cloze (fill-in-the-blank) exercises. Questions can target literal comprehension, vocabulary in context, and inference. Cloze answers accept synonyms and inflections, so the grader is not overly strict on phrasing.
Reading Comprehension is effective from A2 onwards and scales from single-paragraph exercises for beginners to full-article analysis for advanced learners. It pairs well with an optional vocabulary glossary that appears alongside the passage, with pinyin or furigana readings for CJK terms.
Translation
Free-text translation between two languages, graded by an AI rubric across four dimensions: meaning, grammar, vocabulary, and naturalness. Each item can carry multiple reference translations so the grader accepts stylistic variation rather than demanding one exact wording.
Translation is the most effective lesson type for writing practice and grammar consolidation. It is also useful as a warm-up exercise for roleplay, since it forces learners to produce fluent target-language output before they speak.
CJK support out of the box
Language courses for Chinese, Japanese, and Korean learners usually need script annotations that generic platforms do not provide. TutorFlow treats these as first-class authoring options:
| Language | Annotation support |
|---|---|
| Chinese (Mandarin) | Pinyin above characters, tone numbers in vocabulary glossaries |
| Japanese | Furigana (kana readings) above kanji |
| Korean | Hanja annotations when relevant, plus romanization toggles |
You can turn annotations on or off per lesson, which is useful for progressive difficulty. Beginners see pinyin under every character, intermediates only see it for new vocabulary, and advanced learners see none at all.
How a learner experiences a language lesson
Every language practice lesson inside TutorFlow runs in the same shell (header, progress bar, audio controls, and attempt recording) so learners do not need to relearn the interface for each type. The skill-specific differences are in the interaction pattern in the middle:
- Dictation plays audio, shows a text box, and reveals the reference after submission.
- Shadowing plays audio, shows a record button, and compares the transcript to the target after the learner stops recording.
- Roleplay shows the scene briefing, the AI character's opening line, and a reply input with a turn counter.
- Reading Comprehension shows the passage on one side and exercises on the other.
- Translation shows the source sentence and a writing area, with optional "show reference" after submission.
Every attempt is stored with the audio file (for shadowing and roleplay), the transcript, the AI evaluation output, and the time spent. Instructors can revisit any learner's past attempts through the classroom dashboard.
Example course structures
Beginner Korean (absolute beginner to A2)
A short starter course that prioritizes Hangul recognition, core vocabulary, and survival phrases:
- Flashcards: Hangul consonants and vowels.
- Dictation: simple sentences with pinyin-style romanization guides.
- Reading Comprehension: short passages about everyday scenes.
- Shadowing: greetings and polite forms.
- Roleplay: introducing yourself and asking basic questions.
Business English (B1 to B2 professional)
A workplace-focused course for intermediate learners who already speak conversational English:
- Reading Comprehension: short business articles and memos.
- Translation: work emails between the learner's first language and English.
- Roleplay: running a status meeting, negotiating a deadline, giving feedback.
- Flashcards: industry-specific vocabulary (finance, tech, HR).
- Shadowing: presentation pitches and customer service phrases.
JLPT N3 preparation (Japanese learners)
A test-prep course structured around the skills the JLPT actually measures:
- Flashcards: N3 kanji and vocabulary with furigana.
- Reading Comprehension: graded passages with cloze exercises.
- Dictation: news-style sentences with furigana toggles.
- Translation: Japanese to learner's first language for grammar consolidation.
- Shadowing: natural-speed native audio for listening accuracy.
Design principles for effective language lessons
Cover more than one skill per chapter. A chapter that only uses dictation will improve listening but leave speaking and writing untouched. A healthy mix is one listening drill, one speaking drill, one reading or writing drill, and vocabulary support.
Use lenient grading by default. Language learners make small errors constantly. A grader that flags every comma, tone mark, or synonym mismatch kills motivation fast. Start lenient and tighten only for exam-prep contexts.
Keep audio short. For dictation and shadowing, sentences of 30 to 80 characters hit the right difficulty band. Anything longer is a memory test, not a language test.
Give roleplay explicit success criteria. "Have a conversation in French" is a weak scenario. "Order a coffee, ask for the bathroom, and pay with a card" is a scorable one. The more concrete the criteria, the more useful the AI evaluator's feedback.
Show the reference after submission, not before. Learners who see the answer too early skip the productive struggle. Every language practice runner reveals references only after the attempt is submitted.
What is in scope and what is not
TutorFlow language courses cover the core four skills with AI-assisted grading and scalable delivery. A few related capabilities are intentionally out of scope for the current version:
| In scope | Out of scope |
|---|---|
| Dictation, shadowing, roleplay, reading comprehension, translation, vocabulary flashcards | Phoneme-level pronunciation analysis (IPA alignment) |
| Sentence-level transcript comparison for pronunciation scoring | Real-time streaming speech-to-text (all recordings are processed after stop) |
| AI-graded rubrics for translation and roleplay with per-item feedback | Spaced-repetition scheduling (flashcards run linearly) |
| CJK pinyin, furigana, and tone annotations | Learner-to-learner async voice exchange (AI-only conversations) |
| Attempt history with audio, transcript, and evaluation stored per learner | Offline voice processing on the learner's device |
The scope-outs are deliberate. TutorFlow focuses on the language teaching workflow where AI assistance has the highest leverage, rather than rebuilding every specialized tool in the space.