a language by passively turning pages in a textbook.
You actually progress when the language talks again to you.
While you see photos, hear actual sentences, attempt to communicate, and get suggestions, every thing lastly clicks in your head.
Previously, you wanted a instructor by your facet always to get that type of suggestions.
At this time, generative AI can play that position in your cellphone or pc, like an AI language tutor you need to use any time.
After I began studying Mandarin ten years in the past, I noticed many foreigners struggling to be understood by locals in on a regular basis conversations due to poor pronunciation.
It satisfied me that with out good pronunciation, a wealthy vocabulary is ineffective.
I nonetheless bear in mind sitting in my condominium in Shanghai, repeating the identical sentence time and again, with out anybody to right me.
Years later, once I found generative AI, I remembered the engineer in China who was fighting grammar books and tones.
I needed to construct instruments that might have helped me prior to now.
As a startup founder, I would not have a lot free time, so I wanted a option to construct and take a look at new instruments shortly.
That’s the reason I turned to n8n to construct assistants that might have made my Chinese language observe a lot simpler.
On this article, I’ll present how I exploit n8n and multimodal AI to construct a “examine companions” for language studying that:
- Appropriate my pronunciation utilizing Textual content-to-speech capabilities
- Create workouts to review vocabulary lists
- Generate photos for instance phrases or contexts for flash-card model observe
Collectively, they present how AI and low-code platforms like n8n can help anybody studying a posh language.
Even with day by day utilization, all of those collectively price lower than 1 euro per 30 days.
AI For Pronunciation And Oral Comprehension
My title is Samir, a provide chain skilled who struggled with Mandarin throughout his six-year keep in China.
Let me introduce you to Yin, the AI-powered Language coach I developed final week.
It is a net software I designed to help my Chinese language studying journey after greater than 5 years with out practising.
It contains three options:
- Pronunciation Workouts
- A number of Alternative Questions (MCQ)
- Flash Playing cards
I’ll use every characteristic to show how I exploit multimodal AI to enhance my studying comprehension, listening, and pronunciation in Mandarin.
Why is pronunciation in Mandarin so Necessary?
Let me share an actual story from China to focus on the significance of utilizing the proper tone in Mandarin.
Sooner or later, I used to be invited to a job interview on the largest Chinese language specific firm, valued at billions.
The whole dialog was in Chinese language.
I had fastidiously ready my sentences, highlighting how I used information science to enhance warehouse operations.
At one level, I needed to say: “I exploit information science to enhance choosing productiveness within the warehouse.”
The verb “choosing” means taking items from cabinets or racks in a warehouse.
In Chinese language, my colleagues used the verb 拣货 (jiǎn huò) to explain this course of.
However as an alternative of claiming jiǎn huò, I mentioned jiàn huò.
Which is a completely totally different phrase that you just undoubtedly don’t need to use in a job interview.
To maintain it well mannered right here, let’s say jiàn huò is a impolite phrase.
The supervisor burst out laughing.
I didn’t perceive why till I debriefed with the headhunter later and repeated the sentence for her.
That second taught me that pronunciation in Chinese language isn’t nearly sounding pure.
You’ll be able to know hundreds of phrases, but when your tone is mistaken, individuals gained’t perceive you.
For this reason the primary characteristic of my app is an AI pronunciation coach.
Utilizing Speech-to-Textual content Recognition to Practise
Utilizing speech-to-text and reasoning, the app listens to what I say, compares it with the goal sentence, and offers suggestions on which tones or syllables had been off.
The main target right here is on enhancing my pronunciation of logistics and provide chain phrases (my area of experience).
For every phrase, we have now:
- The phrase in Simplified Mandarin Characters: 合同
- The sentence used to practise my pronunciation: 我们需要在发货前签署这份运输合同。
- The English translation: We have to signal this transport contract earlier than transport the products.
For novices, we will even add phonetics (Mandarin pinyin) utilizing the toggle.
How one can observe pronunciation?
I simply need to press the mic button on the backside to document my sentence.
The recording is mechanically despatched to the backend for evaluation that compares my pronunciation with the proper one.
Just a few seconds later, I acquired my suggestions.
The suggestions is kind of detailed; it focuses on the phrases that you just mispronounced.
It’s practically like having a private instructor correcting me in actual time, besides this one by no means will get drained.
After all, this gained’t exchange an important instructor in a one-on-one lesson, however it could possibly aid you to practise after courses.
After I began studying Mandarin, I used to spend evenings (after work) alone, repeating easy sentences to familiarise myself with the nuances of tones.
I didn’t have a suggestions loop on the time; this software would have been very useful.
How does it work?
Textual content-to-speech and reasoning capabilities of GenAI
The backend is a straightforward n8n workflow linked to the frontend by way of a webhook.
The text-to-speech capabilities are used to transcribe the audio file despatched by the entrance finish into phonetics (pinyin).
The output of this Gemini audio transcription node contains the phonetics:
[
{
"content": {
"parts": [
{
"text": "zuò pǐn huò zǒnggòng fàng zài èrshí ge tuōpán shàng.n"
}
],
"position": "mannequin"
},
"finishReason": "STOP",
"avgLogprobs": -0.16858814502584524
}
]This pinyin is then despatched to the AI node Pronounciation Evaluation together with the goal pronunciation.
On this instance, I mispronounced the penultimate phrase.
That is exactly what the agent talked about in his suggestions.
This reveals how we will use text-to-speech capabilities, mixed with the reasoning of generative AI fashions, to enhance our pronunciation.
This may be tailored to any language.
What about picture era and speetch-to-text?
Generative AI for Content material Era
When you observe the person interface of the applying, you discover that every phrase has:
- An illustrative Picture
- A sentence for the context
- Audio transcription out there by way of the microphone icons
This content material is generated utilizing AI fashions to offer quite a lot of educating supplies for the second characteristic: flashcards.
Textual content-to-Speech Options
An effective way to practise pronunciation is to hear and repeat.
Subsequently, earlier than recording my sentence, I can discover ways to pronounce the phrase utilizing this primary speech-to-text characteristic.
For this, I exploit Google’s Textual content-to-Speech API as it’s fairly handy and free.
from gtts import gTTS
def generate_speech(textual content: str, lang: str):
filename = f"{uuid4().hex}.mp3"
filepath = f"./information/gtts/{filename}"
tts = gTTS(textual content=textual content, lang=lang)
tts.save(filepath)With a few strains of code, you possibly can generate the text-to-speech of any phrase utilizing the correct language code.
That is precisely what I used within the software to generate flashcards that I introduced on In the direction of Information Science three years in the past.
The concept on the time was to enhance my listening comprehension by including audio to the flashcard solutions.
What about lengthy sentences?
The issue with Google Textual content-to-speech is the robotic voice.
Fortuitously, we have now eleven labs.
The workflow above is linked to the app by way of webhook.
The Eleven labs node that takes the output of the AI Agent Generate Instance to generate the audio model of the sentence.
The person can now take heed to the sentence pronounced “like” a local speaker.
What’s remaining? Questions and illustrations …
Educating materials era
As defined within the earlier part, the sentences are additionally generated utilizing AI.
The AI Agent node, powered by Gemini, takes the phrase to review as enter and makes use of the system immediate beneath to generate a sentence.
You're a Chinese language language tutor for professionals.
Given a Chinese language phrase, you MUST return a JSON object with EXACTLY these keys:
- "sentence": a brief Chinese language sentence utilizing the phrase in a enterprise or
daily-life context
- "pinyin": the pinyin of the total sentence
- "english": the English translation of the sentence
Return ONLY legitimate JSON. No explanations, no backticks, no additional textual content.
Instance:
{
"sentence": "我去仓库检查货物。",
"pinyin": "Wǒ qù cāngkù jiǎnchá huòwù.",
"english": "I am going to the warehouse to examine the products."
}That ensures an almost infinite number of workouts.
And the cherry on the cake is the picture generated with Gemini’s Nano Banana to assist us join a phrase to its context.
After studying hundreds of Chinese language characters, I seen that photos assist with memorising new phrases.
That is exactly what I exploit within the flashcards characteristic.
The n8n backend gives to the front-end:
- The phrase in Chinese language that you just need to study with pinyin and English translation
- An instance sentence and its translation generated by GPT
- An illustrative picture generated by Gemini
The entrance finish then manages the card-flipping mechanism.
If you wish to recreate this answer tailor-made to your wants, I’ve shared the same workflow on my GitHub.
Do you want multiple-choices questions? Gen AI might help!
Generate Workouts from a vocabulary record
For the final characteristic, we generate multiple-choice inquiries to study the identical vocabulary record.
We ask Gemini to generate questions from the vocabulary record, utilizing multiple-choice choices with just one right reply.
[
{
"output": {
"question": "Which of the following is the correct Chinese translation for 'Variable Pricing'? Please answer with A, B, C, or D.",
"options": {
"A": "仓库",
"B": "可变定价",
"C": "卡车司机",
"D": "投标"
},
"correct": "B",
"right_feedback": "Great job! 可变定价 (kě biàn dìng jià) means Variable Pricing.",
"wrong_feedback": "Oops! The correct answer is B: 可变定价 (kě biàn dìng jià), which means Variable Pricing."
}
}
]The front-end makes use of this output to offer the questions with tailored suggestions.
The backend of this characteristic relies on an n8n workflow that I additionally shared on my GitHub: AI-Powered Language Trainer utilizing GPT.
Conclusion
I developed this app to experiment with how AI may improve my studying capabilities.
After practically 5 years with out talking Chinese language, this multimodal AI assistant has confirmed to be an important assist.
The whole backend is constructed on n8n for fast prototyping and seamless integration.
You aren’t conversant in n8n and need to study?
I’ve an entire tutorial, designed for novices, on my YouTube channel that may information you from occasion creation to credential setup.
After this tutorial, it is possible for you to to make use of any of the workflows shared in my repository.
As I would not have time to decide to in-person Chinese language courses, I can have an assistant who will adapt to my schedule.
Can we do higher?
On the “roadmap” of this small facet undertaking, I’ve:
- Including advanced grammar workouts that may very well be accomplished orally (combining studying comprehension, grammar and pronunciation)
- Implementing a writing module that might right my calligraphy utilizing picture processing
Relying on my availability, I’ll intention to ship it by Q1 2026.
About Me
Let’s join on LinkedIn and Twitter; I’m a Provide Chain Engineer utilizing information analytics to enhance logistics operations and cut back prices.
For consulting or recommendation on analytics and sustainable provide chain transformation, please contact me by way of Logigreen Consulting.
