Update app.py

app.py

@@ -372,6 +372,22 @@ def run_prompts():
     prompt_form = st.empty()
     with prompt_form.form("prompt_form"):
         #prompt = st.text_input("Token Name")
+        st.caption(
+            """
+            In general, the best stable diffusion prompts will have this form:
+
+            `A [prompt], [prompt_keywords]`
+
+            The `[prompt]`can be anything you’re thinking of, but Stable Diffusion still struggles with compositionality, so it shouldn’t be more than one or two main things (say, a beaver wearing a suit, or a cat samurai with a pet pug). The main subject should be mostly composed of adjectives and nouns. Avoid verbs, as Stable Diffusion has a hard time interpreting them correctly.
+
+            `[Prompt Keywords]` can be anything you want to condition the image on. I wouldn’t add too many, maybe only 1 to 3. These can really vary a lot but some good ones are: `[concept art, steampunk, trending in artstation, good composition, hyper realistic, oil on canvas, vivid colors]`.
+
+            Additionally, adding the name of an artist as a cue will make the picture look like something that artist made, though it may condition the image’s contents, especially if that artist had narrow themes (Beatrix Potter gets you spurious rabbits, for instance).
+
+            You can also find many great prompts if you use the Lexica.art prompt search engine to find images you like and make tweaks to their prompts.
+            """
+        )
+        st.text("")
         full_prompt = st.text_input("Prompt")
         prompt_keywords = st.text_input(f"Prompt Keywords")
         submitted = st.form_submit_button(f"Submit")