diff --git a/README.md b/README.md
index f55aefc..e8e9e95 100644
--- a/README.md
+++ b/README.md
@@ -10,10 +10,14 @@ Station Announcer is a small Flask app that keeps your Immich library and Mastod
    IMMICH_API_URL=https://photos.4angle.com/
    IMMICH_API_KEY=your-immich-api-key
    OPENAI_API_KEY=your-openai-api-key
+   OPENAI_MODEL=gpt-4o-mini
    MASTODON_BASE_URL=http://localhost:3000
    MASTODON_ACCESS_TOKEN=your-mastodon-token
    ```
 
+   `OPENAI_MODEL` is optional; set it to override the default model used for both
+   alt-text generation and copy refinement.
+
 2. Install dependencies and run the development server:
 
    ```bash
diff --git a/station_announcer/__init__.py b/station_announcer/__init__.py
index f05ada8..ace6bfc 100644
--- a/station_announcer/__init__.py
+++ b/station_announcer/__init__.py
@@ -8,7 +8,7 @@ from flask import Flask
 sys.path.append("/home/edward/src/2024/UniAuth")  # isort:skip
 
 from .cache import AltTextCache
-from .config import load_settings
+from .config import DEFAULT_OPENAI_MODEL, load_settings
 from .immich import ImmichClient
 from .mastodon import MastodonClient
 from .openai_client import AltTextGenerator
@@ -60,7 +60,7 @@ def create_app() -> Flask:
     app.alt_text_cache = AltTextCache(db_path)
     app.alt_text_generator = AltTextGenerator(
         api_key=app.config["OPENAI_API_KEY"],
-        model=app.config.get("OPENAI_MODEL", "gpt-4o-mini"),
+        model=app.config.get("OPENAI_MODEL", DEFAULT_OPENAI_MODEL),
     )
     mastodon_token = app.config.get("MASTODON_ACCESS_TOKEN")
     if mastodon_token:
diff --git a/station_announcer/config.py b/station_announcer/config.py
index 9abb2b2..dc82c59 100644
--- a/station_announcer/config.py
+++ b/station_announcer/config.py
@@ -8,6 +8,7 @@ from typing import Dict
 from dotenv import load_dotenv
 
 DEFAULT_IMMICH_URL = "https://photos.4angle.com/"
+DEFAULT_OPENAI_MODEL = "gpt-4o-mini"
 
 
 class ConfigError(RuntimeError):
@@ -26,7 +27,7 @@ def load_settings() -> Dict[str, str]:
         "IMMICH_API_KEY": os.getenv("IMMICH_API_KEY", ""),
         "OPENAI_API_KEY": os.getenv("OPENAI_API_KEY", ""),
         "RECENT_DAYS": int(os.getenv("RECENT_DAYS", "3")),
-        "OPENAI_MODEL": os.getenv("OPENAI_MODEL", "gpt-4o-mini"),
+        "OPENAI_MODEL": os.getenv("OPENAI_MODEL", DEFAULT_OPENAI_MODEL),
         "STATION_DB": db_path,
         "SECRET_KEY": os.getenv("SECRET_KEY", ""),
         "MASTODON_BASE_URL": os.getenv(
diff --git a/station_announcer/openai_client.py b/station_announcer/openai_client.py
index 3e68a28..4474f6b 100644
--- a/station_announcer/openai_client.py
+++ b/station_announcer/openai_client.py
@@ -5,6 +5,7 @@ from __future__ import annotations
 import requests
 import typing
 
+from .config import DEFAULT_OPENAI_MODEL
 
 class OpenAIClientError(RuntimeError):
     """Raised when the OpenAI API cannot fulfill a request."""
@@ -21,7 +22,7 @@ class TextImprovementError(OpenAIClientError):
 class AltTextGenerator:
     """Request alt text from a GPT compatible OpenAI endpoint."""
 
-    def __init__(self, api_key: str, model: str = "gpt-4.1") -> None:
+    def __init__(self, api_key: str, model: str = DEFAULT_OPENAI_MODEL) -> None:
         if not api_key:
             raise ValueError("OPENAI_API_KEY is required")
         self.model = model
@@ -106,6 +107,7 @@ class AltTextGenerator:
         instructions: str | None = None,
         hashtag_counts: str | None = None,
         thread_context: list[str] | None = None,
+        alt_texts: list[str] | None = None,
     ) -> str:
         if not draft_text or not draft_text.strip():
             raise TextImprovementError("Post text cannot be empty")
@@ -150,6 +152,16 @@ class AltTextGenerator:
                 f"{user_content}\n\nHashtag history (tag with past uses):\n"
                 f"{hashtag_counts.strip()}"
             )
+        if alt_texts:
+            cleaned_alt_texts = [entry.strip() for entry in alt_texts if entry.strip()]
+            if cleaned_alt_texts:
+                alt_text_block = "\n".join(
+                    f"{index + 1}. {value}"
+                    for index, value in enumerate(cleaned_alt_texts)
+                )
+                user_content = (
+                    f"{user_content}\n\nAlt text for attached images:\n{alt_text_block}"
+                )
 
         payload = {
             "model": self.model,
diff --git a/station_announcer/routes.py b/station_announcer/routes.py
index 948862d..c38c3f0 100644
--- a/station_announcer/routes.py
+++ b/station_announcer/routes.py
@@ -397,9 +397,18 @@ def compose_draft():
                     context_entries.append(latest_text)
                 if context_entries:
                     thread_context = context_entries
+            alt_texts = [
+                entry["alt_text"].strip()
+                for entry in asset_entries
+                if entry.get("alt_text")
+            ]
             try:
                 post_text = generator.improve_post_text(
-                    post_text, instructions, hashtag_counts, thread_context
+                    post_text,
+                    instructions,
+                    hashtag_counts,
+                    thread_context,
+                    alt_texts,
                 )
                 flash("Post refined with ChatGPT.")
             except TextImprovementError as exc: