Translator App using Python, Streamlit, and Deep Translator

Introduction

This project walks you through building a multi-language translator app with Python, Streamlit, and the deep-translator library. You can translate text input — or upload an entire .txt document — across 100+ languages using Google Translate under the hood. Translation pairs naturally with text understanding — see the companion sentiment analysis app for the NLP side.

Prerequisites

Python 3.8+ — python.org

Libraries:

code

pip install streamlit deep-translator

Step 1: Create the Script

Create translator_app.py and paste the following:

code

import streamlit as st
from deep_translator import GoogleTranslator

st.set_page_config(page_title="Translator App", page_icon="🌍")
st.title("🌍 Multi-Language Translator")

LANGUAGES = GoogleTranslator().get_supported_languages(as_dict=True)
lang_names = list(LANGUAGES.keys())

col1, col2 = st.columns(2)
with col1:
    source_lang = st.selectbox("Source Language", ["auto"] + lang_names)
with col2:
    target_lang = st.selectbox("Target Language", lang_names,
                               index=lang_names.index("english"))

tab1, tab2 = st.tabs(["Translate Text", "Translate Document"])

with tab1:
    text_input = st.text_area("Enter text to translate:", height=150)
    if st.button("Translate Text") and text_input:
        src = "auto" if source_lang == "auto" else LANGUAGES[source_lang]
        tgt = LANGUAGES[target_lang]
        result = GoogleTranslator(source=src, target=tgt).translate(text_input)
        st.subheader("Translation:")
        st.success(result)

with tab2:
    uploaded = st.file_uploader("Upload a .txt file", type=["txt"])
    if st.button("Translate Document") and uploaded:
        content = uploaded.read().decode("utf-8")
        src = "auto" if source_lang == "auto" else LANGUAGES[source_lang]
        tgt = LANGUAGES[target_lang]
        result = GoogleTranslator(source=src, target=tgt).translate(content[:5000])
        st.text_area("Translated Document:", result, height=300)
        st.download_button("Download Translation", result, file_name="translated.txt")

Step 2: Run the App

code

streamlit run translator_app.py

Step 3: Use the App

Pick source and target languages from the dropdowns.

Use the Translate Text tab for quick text translation.

Use the Translate Document tab to upload and translate a .txt file.

Click Download Translation to save the output.

How It Works

The GoogleTranslator class from deep-translator does the heavy lifting: you specify a source and target language code, call .translate(text), and get a string back. The library handles the HTTP call to Google's endpoint, chunking of long text, and retries — which is why it is the go-to choice over raw requests.

The Streamlit layer is two st.selectbox widgets for languages (populated from GoogleTranslator().get_supported_languages()), a text area, and a button. Gating the translation behind st.button("Translate") is deliberate: without it, every keystroke-triggered rerun would fire an API call, which is slow and can get rate-limited.

Language codes are ISO 639-1 strings ("en", "hi", "fr"). Mapping friendly names to codes with a small dictionary keeps the UI clean and the API calls valid.

Key Concepts

`GoogleTranslator` — Wraps Google Translate with a clean Python API.

`get_supported_languages()` — Returns a dict of all supported language names and codes.

`st.tabs()` — Creates a tabbed interface separating text and document modes.

`st.download_button()` — Lets users save the translated content as a file.

What to Try Next

Support PDF and DOCX files using PyMuPDF or python-docx.

Show a detected source language label when using the auto-detect mode.

Add a character counter to display the remaining translation limit.

Common Errors & Fixes

NoTranslationError / empty output — the source text is blank or only punctuation. Validate with if not text.strip(): st.warning(...) before calling the translator.

Invalid destination language — you passed a display name ("Hindi") where a code ("hi") is expected. Build the name→code map once at startup.

Random network failures — Google's endpoint occasionally throttles. Wrap the call in try/except and retry once after a short sleep.

Very long text gets cut — translate in chunks of ~4500 characters and join the results.

Translation returns the original text — source and target language codes are identical (both "en"); check the selectbox defaults.

App hangs on submit — the network call runs without feedback; wrap it in with st.spinner("Translating..."): so the UI stays responsive while waiting.

FAQ

Is deep-translator free?

Yes — it uses the free Google endpoint without an API key. For production volume, switch to the official Cloud Translation API.

Can I auto-detect the source language?

Yes — pass source="auto" and Google detects it per request.

How do I translate a whole file?

Read the file, split into paragraphs, translate each, and join — or render a side-by-side view with st.columns(2), a pattern shared with the quiz app's layout.

Does it work offline?

No — deep-translator calls Google's endpoint over the network. For offline translation you would need a local model like Argos Translate.