model default to default_model for mlx/gguf models, and add new text as optional layer; allow --no-original option to remove original content in pdf, just save the translated content

deep_translator/chatgpt.py

@@ -58,8 +58,10 @@ class ChatGptTranslator(BaseTranslator):
         prompt = f"Translate the text below into {self.target}.\n"
         prompt += f'Text: "{text}"'
 
+        # if model is empty (for mlx_lm.server, the model should be default_model)
+        # export OPENAI_MODEL=default_model
         response = client.chat.completions.create(
-            model=self.model,
+            model=self.model if self.model else "default_model",
             messages=[
                 {
                     "role": "user",
@@ -67,6 +69,7 @@ class ChatGptTranslator(BaseTranslator):
                 }
             ],
         )
+        
 
         return response.choices[0].message.content
 

docs/README.rst

@@ -184,7 +184,7 @@ or even directly from terminal:
 
     or shorter
 
-    $ dt -tg de -txt "hello world"
+    $ dt --translator chatgpt -tg de -txt "hello world"
 
 
 =====
@@ -594,7 +594,7 @@ ChatGpt Translator
 
     You can provide your api key, api base as an argument or you can export it as an env var
     e.g. 
-    
+
     `export OPENAI_API_KEY="your_key"`
 
     `export OPENAI_API_BASE=https://api.openai.com/v1`

translator_cli.py

@@ -11,7 +11,8 @@ TRANSLATORS = {
     'chatgpt': ChatGptTranslator,
 }
 
-def translate_pdf(input_file: str, source_lang: str, target_lang: str, layer: str = "Korean", translator_name: str = "google"):
+def translate_pdf(input_file: str, source_lang: str, target_lang: str, layer: str = "Text", 
+                 translator_name: str = "google", text_color: str = "darkred", keep_original: bool = True):
     """
     Translate a PDF file from source language to target language
     
@@ -19,11 +20,25 @@ def translate_pdf(input_file: str, source_lang: str, target_lang: str, layer: st
         input_file: Path to input PDF file
         source_lang: Source language code (e.g. 'en', 'fr')
         target_lang: Target language code (e.g. 'ko', 'ja') 
-        layer: Name of the OCG layer (default: "Korean")
+        layer: Name of the OCG layer (default: "Text")
         translator_name: Name of the translator to use (default: "google")
+        text_color: Color of translated text (default: "darkred")
+        keep_original: Whether to keep original text visible (default: True)
     """
-    # Define color "white"
+    # Define colors
     WHITE = pymupdf.pdfcolor["white"]
+    
+    # Color mapping
+    COLOR_MAP = {
+        "darkred": (0.8, 0, 0),
+        "black": (0, 0, 0),
+        "blue": (0, 0, 0.8),
+        "darkgreen": (0, 0.5, 0),
+        "purple": (0.5, 0, 0.5),
+    }
+    
+    # Get RGB color values, default to darkred if color not found
+    rgb_color = COLOR_MAP.get(text_color.lower(), COLOR_MAP["darkred"])
 
     # This flag ensures that text will be dehyphenated after extraction.
     textflags = pymupdf.TEXT_DEHYPHENATE
@@ -43,9 +58,12 @@ def translate_pdf(input_file: str, source_lang: str, target_lang: str, layer: st
     # Open the document
     doc = pymupdf.open(input_file)
 
-    # Define an Optional Content layer in the document.
-    # Activate it by default.
-    ocg_xref = doc.add_ocg(layer, on=True)
+    # Define an Optional Content layer for translation
+    ocg_trans = doc.add_ocg(layer, on=True)
+    
+    # If not keeping original, create a layer for original text and hide it
+    if not keep_original:
+        ocg_orig = doc.add_ocg("Original", on=False)
 
     # Iterate over all pages
     for page in doc:
@@ -60,12 +78,26 @@ def translate_pdf(input_file: str, source_lang: str, target_lang: str, layer: st
             # Invoke the actual translation
             translated = translator.translate(text)
 
-            # Cover the source text with a white rectangle.
-            page.draw_rect(bbox, color=None, fill=WHITE, oc=ocg_xref)
-
-            # Write the translated text into the original rectangle
+            if not keep_original:
+                # Move original text to hidden layer
+                page.insert_htmlbox(
+                    bbox,
+                    text,
+                    css="* {font-family: sans-serif;}",
+                    oc=ocg_orig
+                )
+                # Clear original text area in base layer
+                page.draw_rect(bbox, color=None, fill=WHITE)
+            else:
+                # Cover the original text only in translation layer
+                page.draw_rect(bbox, color=None, fill=WHITE, oc=ocg_trans)
+
+            # Write the translated text in specified color
             page.insert_htmlbox(
-                bbox, translated, css="* {font-family: sans-serif;}", oc=ocg_xref
+                bbox,
+                translated,
+                css=f"* {{font-family: sans-serif; color: rgb({int(rgb_color[0]*255)}, {int(rgb_color[1]*255)}, {int(rgb_color[2]*255)});}}",
+                oc=ocg_trans
             )
 
     doc.subset_fonts()
@@ -74,26 +106,51 @@ def translate_pdf(input_file: str, source_lang: str, target_lang: str, layer: st
 
 def main():
     """
-      can be invoked like this:
-      python translator_cli.py --source english --target zh-CN "/Users/david/Downloads/Level_up_coding_by_ai.pdf"
+    can be invoked like this:
+    ```
+    # Basic usage
+    python translator_cli.py --source english --target zh-CN input.pdf
+
+    # With custom color and hiding original text
+    python translator_cli.py --source english --target zh-CN --color blue --no-original input.pdf
+
+    # Using ChatGPT translator
+    export OPENAI_API_KEY=sk-proj-xxxx
+    export OPENAI_API_BASE=https://api.xxxx.com/v1
+    export OPENAI_API_BASE=http://localhost:8080/v1 #  for local llm api
+    python translator_cli.py --source english --translator chatgpt --target zh-CN input.pdf
+    
+    # do not keep original text as an optional layer:
+    python translator_cli.py --source english --translator chatgpt --target zh-CN --no-original input.pdf
+    
+    ```
+
+    The translated content is an optional content layer in the new PDF file. 
+    The optional layer can be hidden in Acrobat PDF Reader and Foxit Reader.
     """
     
     parser = argparse.ArgumentParser(description='Translate PDF documents.')
     parser.add_argument('input_file', help='Input PDF file path')
     parser.add_argument('--source', '-s', default='en',
                        help='Source language code (default: en)')
-    parser.add_argument('--target', '-t', default='ko',
-                       help='Target language code (default: ko)')
-    parser.add_argument('--layer', '-l', default='Korean',
-                       help='Name of the OCG layer (default: Korean)')
+    parser.add_argument('--target', '-t', default='zh-CN',
+                       help='Target language code (default: zh-CN)')
+    parser.add_argument('--layer', '-l', default='Text',
+                       help='Name of the OCG layer (default: Text)')
     parser.add_argument('--translator', '-tr', default='google',
                        choices=list(TRANSLATORS.keys()),
                        help='Translator to use (default: google)')
+    parser.add_argument('--color', '-c', default='darkred',
+                       choices=['darkred', 'black', 'blue', 'darkgreen', 'purple'],
+                       help='Color of translated text (default: darkred)')
+    parser.add_argument('--no-original', action='store_true',
+                       help='Do not keep original text in base layer (default: False)')
 
     args = parser.parse_args()
 
     try:
-        translate_pdf(args.input_file, args.source, args.target, args.layer, args.translator)
+        translate_pdf(args.input_file, args.source, args.target, args.layer, 
+                     args.translator, args.color, not args.no_original)
     except Exception as e:
         print(f"Error: {str(e)}")
         exit(1)