[Python] Why is the translated image showing weird characters?

Alibi is a full-stack offline web app that helps immigrants translate legal/medical documents and ask questions in their native language. It runs fully locally — no cloud APIs, no user accounts, no internet connection required.

Image Translator: Uses pytesseract to extract English text from images, translates it using libretranslate, then rebuilds the image with translated text. Tech stack: Frontend: HTML/CSS/JS or React Backend: Flask OCR: pytesseract Translation: libretranslate (self-hosted, using docker volume with downloaded language packs/models);

This is the image_utils.py methods:

from PIL import Image, ImageDraw, ImageFont
import pytesseract
import os
from document_translator import translator
from collections import defaultdict

# Set Tesseract path (adjust if needed)
pytesseract.pytesseract.tesseract_cmd = r"C:\Program Files\Tesseract-OCR\tesseract.exe"
def image_to_string(image_path):
text = pytesseract.image_to_string(Image.open(image_path))
print("[OCR OUTPUT]", text)
return text

# Extracts readable English text from an image
def font_algorithm(lang_code: str, font_base_dir="C:/Fonts/Noto", default_font="arial.ttf") -> str:
"""
Returns the best available font path for a given language code using internal font map.
Falls back to a default system font if not found.
"""
LANGUAGE_FONT_MAP = {
'en': 'NotoSans-Regular.ttf',
'aa': 'NotoSans-Regular.ttf',
'ab': 'NotoSans-Regular.ttf',
'ae': 'NotoSans-Regular.ttf',
'af': 'NotoSans-Regular.ttf',
'ak': 'NotoSans-Regular.ttf',
'am': 'NotoSansEthiopic-Regular.ttf',
'an': 'NotoSans-Regular.ttf',
'ar': 'NotoNaskhArabic-Regular.ttf',
'as': 'NotoSansBengali-Regular.ttf',
'av': 'NotoSans-Regular.ttf',
'ay': 'NotoSans-Regular.ttf',
'az': 'NotoSans-Regular.ttf',
'ba': 'NotoSans-Regular.ttf',
'be': 'NotoSans-Regular.ttf',
'bg': 'NotoSans-Regular.ttf',
'bh': 'NotoSansDevanagari-Regular.ttf',
'bi': 'NotoSans-Regular.ttf',
'bm': 'NotoSans-Regular.ttf',
'bn': 'NotoSansBengali-Regular.ttf',
'bo': 'NotoSansTibetan-Regular.ttf',
'br': 'NotoSans-Regular.ttf',
'bs': 'NotoSans-Regular.ttf',
'ca': 'NotoSans-Regular.ttf',
'ce': 'NotoSans-Regular.ttf',
'ch': 'NotoSans-Regular.ttf',
'co': 'NotoSans-Regular.ttf',
'cr': 'NotoSansCanadianAboriginal-Regular.ttf',
'cs': 'NotoSans-Regular.ttf',
'cu': 'NotoSans-Regular.ttf',
'cv': 'NotoSans-Regular.ttf',
'cy': 'NotoSans-Regular.ttf',
'da': 'NotoSans-Regular.ttf',
'de': 'NotoSans-Regular.ttf',
'dv': 'NotoSansThaana-Regular.ttf',
'dz': 'NotoSansTibetan-Regular.ttf',
'ee': 'NotoSans-Regular.ttf',
'el': 'NotoSansGreek-Regular.ttf',
'eo': 'NotoSans-Regular.ttf',
'es': 'NotoSans-Regular.ttf',
'et': 'NotoSans-Regular.ttf',
'eu': 'NotoSans-Regular.ttf',
'fa': 'NotoNaskhArabic-Regular.ttf',
'ff': 'NotoSans-Regular.ttf',
'fi': 'NotoSans-Regular.ttf',
'fj': 'NotoSans-Regular.ttf',
'fo': 'NotoSans-Regular.ttf',
'fr': 'NotoSans-Regular.ttf',
'fy': 'NotoSans-Regular.ttf',
'ga': 'NotoSans-Regular.ttf',
'gd': 'NotoSans-Regular.ttf',
'gl': 'NotoSans-Regular.ttf',
'gn': 'NotoSans-Regular.ttf',
'gu': 'NotoSansGujarati-Regular.ttf',
'gv': 'NotoSans-Regular.ttf',
'ha': 'NotoSans-Regular.ttf',
'he': 'NotoSansHebrew-Regular.ttf',
'hi': 'NotoSansDevanagari-Regular.ttf',
'ho': 'NotoSans-Regular.ttf',
'hr': 'NotoSans-Regular.ttf',
'ht': 'NotoSans-Regular.ttf',
'hu': 'NotoSans-Regular.ttf',
'hy': 'NotoSansArmenian-Regular.ttf',
'hz': 'NotoSans-Regular.ttf',
'ia': 'NotoSans-Regular.ttf',
'id': 'NotoSans-Regular.ttf',
'ie': 'NotoSans-Regular.ttf',
'ig': 'NotoSans-Regular.ttf',
'ii': 'NotoSansYi-Regular.ttf',
'ik': 'NotoSans-Regular.ttf',
'io': 'NotoSans-Regular.ttf',
'is': 'NotoSans-Regular.ttf',
'it': 'NotoSans-Regular.ttf',
'iu': 'NotoSansCanadianAboriginal-Regular.ttf',
'ja': 'NotoSansJP-Regular.otf',
'jv': 'NotoSans-Regular.ttf',
'ka': 'NotoSansGeorgian-Regular.ttf',
'kg': 'NotoSans-Regular.ttf',
'ki': 'NotoSans-Regular.ttf',
'kj': 'NotoSans-Regular.ttf',
'kk': 'NotoSans-Regular.ttf',
'kl': 'NotoSans-Regular.ttf',
'km': 'NotoSansKhmer-Regular.ttf',
'kn': 'NotoSansKannada-Regular.ttf',
'ko': 'NotoSansKR-Regular.otf',
'kr': 'NotoSans-Regular.ttf',
'ks': 'NotoSans-Regular.ttf',
'ku': 'NotoSans-Regular.ttf',
'kv': 'NotoSans-Regular.ttf',
'kw': 'NotoSans-Regular.ttf',
'ky': 'NotoSans-Regular.ttf',
'la': 'NotoSans-Regular.ttf',
'lb': 'NotoSans-Regular.ttf',
'lg': 'NotoSans-Regular.ttf',
'li': 'NotoSans-Regular.ttf',
'ln': 'NotoSans-Regular.ttf',
'lo': 'NotoSansLao-Regular.ttf',
'lt': 'NotoSans-Regular.ttf',
'lu': 'NotoSans-Regular.ttf',
'lv': 'NotoSans-Regular.ttf',
'mg': 'NotoSans-Regular.ttf',
'mh': 'NotoSans-Regular.ttf',
'mi': 'NotoSans-Regular.ttf',
'mk': 'NotoSans-Regular.ttf',
'ml': 'NotoSansMalayalam-Regular.ttf',
'mn': 'NotoSansMongolian-Regular.ttf',
'mr': 'NotoSansDevanagari-Regular.ttf',
'ms': 'NotoSans-Regular.ttf',
'mt': 'NotoSans-Regular.ttf',
'my': 'NotoSansMyanmar-Regular.ttf',
'na': 'NotoSans-Regular.ttf',
'nb': 'NotoSans-Regular.ttf',
'nd': 'NotoSans-Regular.ttf',
'ne': 'NotoSansDevanagari-Regular.ttf',
'ng': 'NotoSans-Regular.ttf',
'nl': 'NotoSans-Regular.ttf',
'nn': 'NotoSans-Regular.ttf',
'no': 'NotoSans-Regular.ttf',
'nr': 'NotoSans-Regular.ttf',
'nv': 'NotoSans-Regular.ttf',
'ny': 'NotoSans-Regular.ttf',
'oc': 'NotoSans-Regular.ttf',
'oj': 'NotoSansCanadianAboriginal-Regular.ttf',
'om': 'NotoSans-Regular.ttf',
'or': 'NotoSansOriya-Regular.ttf',
'os': 'NotoSans-Regular.ttf',
'pa': 'NotoSansGurmukhi-Regular.ttf',
'pi': 'NotoSansDevanagari-Regular.ttf',
'pl': 'NotoSans-Regular.ttf',
'ps': 'NotoNaskhArabic-Regular.ttf',
'pt': 'NotoSans-Regular.ttf',
'qu': 'NotoSans-Regular.ttf',
'rm': 'NotoSans-Regular.ttf',
'rn': 'NotoSans-Regular.ttf',
'ro': 'NotoSans-Regular.ttf',
'ru': 'NotoSans-Regular.ttf',
'rw': 'NotoSans-Regular.ttf',
'sa': 'NotoSansDevanagari-Regular.ttf',
'sc': 'NotoSans-Regular.ttf',
'sd': 'NotoSansDevanagari-Regular.ttf',
'se': 'NotoSans-Regular.ttf',
'sg': 'NotoSans-Regular.ttf',
'si': 'NotoSansSinhala-Regular.ttf',
'sk': 'NotoSans-Regular.ttf',
'sl': 'NotoSans-Regular.ttf',
'sm': 'NotoSans-Regular.ttf',
'sn': 'NotoSans-Regular.ttf',
'so': 'NotoSans-Regular.ttf',
'sq': 'NotoSans-Regular.ttf',
'sr': 'NotoSans-Regular.ttf',
'ss': 'NotoSans-Regular.ttf',
'st': 'NotoSans-Regular.ttf',
'su': 'NotoSans-Regular.ttf',
'sv': 'NotoSans-Regular.ttf',
'sw': 'NotoSans-Regular.ttf',
'ta': 'NotoSansTamil-Regular.ttf',
'te': 'NotoSansTelugu-Regular.ttf',
'tg': 'NotoSans-Regular.ttf',
'th': 'NotoSansThai-Regular.ttf',
'ti': 'NotoSansEthiopic-Regular.ttf',
'tk': 'NotoSans-Regular.ttf',
'tl': 'NotoSans-Regular.ttf',
'tn': 'NotoSans-Regular.ttf',
'to': 'NotoSans-Regular.ttf',
'tr': 'NotoSans-Regular.ttf',
'ts': 'NotoSans-Regular.ttf',
'tt': 'NotoSans-Regular.ttf',
'tw': 'NotoSans-Regular.ttf',
'ty': 'NotoSans-Regular.ttf',
'ug': 'NotoSansArabic-Regular.ttf',
'uk': 'NotoSans-Regular.ttf',
'ur': 'NotoNastaliqUrdu-Regular.ttf',
'uz': 'NotoSans-Regular.ttf',
've': 'NotoSans-Regular.ttf',
'vi': 'NotoSans-Regular.ttf',
'vo': 'NotoSans-Regular.ttf',
'wa': 'NotoSans-Regular.ttf',
'wo': 'NotoSans-Regular.ttf',
'xh': 'NotoSans-Regular.ttf',
'yi': 'NotoSansHebrew-Regular.ttf',
'yo': 'NotoSans-Regular.ttf',
'za': 'NotoSans-Regular.ttf',
'zh': 'NotoSansSC-Regular.otf',
'zu': 'NotoSans-Regular.ttf'
}

font_file = LANGUAGE_FONT_MAP.get(lang_code, LANGUAGE_FONT_MAP['en']) # fallback to English
font_path = os.path.join(font_base_dir, font_file)

if os.path.isfile(font_path):
return font_path

print(f"[font_algorithm] Font not found on disk for language '{lang_code}': {font_file}. Using fallback: {default_font}")
return default_font

# Overlays translated text on image and saves it
# Overlays translated text on image and saves it
def overlay_translated_text_on_image(original_image_path: str, lang: str, output_path: str):
image = Image.open(original_image_path)
draw = ImageDraw.Draw(image)

# Dynamically get best font path using the algorithm
font_path = font_algorithm(lang) # uses the lang argument
try:
font = ImageFont.truetype(font_path, size=24)
except OSError:
print(f"[WARNING] Font file not found or invalid: {font_path}. Falling back to Arial.")
font = ImageFont.truetype("arial.ttf", size=24)

data = pytesseract.image_to_data(image, output_type=pytesseract.Output.DICT)

lines = defaultdict(list)
for i in range(len(data["text"])):
word = data["text"]
conf = int(data["conf"])

if word.strip() == "" or conf < 0:
continue

key = (data["block_num"], data["line_num"])
word_info = {
"text": word,
"left": data["left"],
"top": data["top"],
"width": data["width"],
"height": data["height"]
}
lines[key].append(word_info)

# Translate and draw each line
for line_words in lines.values():
english_line = " ".join([w["text"] for w in line_words])
if not english_line.strip():
continue

print("[TRANSLATE INPUT]", english_line)
translated_line = translator.translate(english_line, lang)
print("[TRANSLATE OUTPUT]", translated_line)

# Position to draw: left/top of first word
x = line_words[0]["left"]
y = line_words[0]["top"]

# Erase original text area with white box
text_width = sum(w["width"] for w in line_words)
text_height = max(w["height"] for w in line_words)
draw.rectangle([x, y, x + text_width, y + text_height], fill="white")

# Draw translated text
draw.text((x, y), translated_line, fill="black", font=font)

image.save(output_path)


This is the translator.py:

import requests

def translate(text: str, target_lang: str) -> str:
print("[TRANSLATE INPUT]", text)

response = requests.post(
"http://localhost:5000/translate",
data={
"q": text,
"source": "en",
"target": target_lang,
"format": "text"
}
)

if response.status_code == 200:
translated = response.json().get("translatedText", "")
print("[TRANSLATE OUTPUT]", translated)
return translated
else:
print("[TRANSLATE ERROR]", response.status_code, response.text)
raise Exception(f"Translation failed: {response.status_code} {response.text}")


Image to be translated to Bengali

Image after running through translation algorithm

Continue reading...