Upload character_sheet_prompt.md with huggingface_hub

character_sheet_prompt.md

@@ -15,14 +15,14 @@ A single image, 4 panels arranged left to right, all on a clean white background
 
 The same person's exact face, hair, and clothing are preserved across all four panels.
 
-## Model
+> ⚠️ **The reference image resolution must be `1536×1024`.** The LTX character-sheet LoRA was
+> trained on sheets at this exact size — ask for `1536x1024` / landscape when generating it, and
+> use `ref_resize_mode = native_resolution` on the ComfyUI node at inference time (see below).
 
-`gpt-image-2`, via the OpenAI Images **edit** endpoint (`client.images.edit`), so the input photo
-is used as the base to edit rather than just a text-to-image reference.
+## Model
 
-- `size`: `"1536x1024"`
-- `quality`: `"low"` is enough for training data (cheaper, ~$0.011/image); use `"medium"`/`"high"`
-  for a single polished reference at inference time.
+`gpt-image-2` — the easiest way is to just **upload your photo to ChatGPT** and paste the prompt
+below (image generation/editing is built in, no API needed). Ask for it at **1536×1024** / landscape.
 
 ## The prompt
 
@@ -59,34 +59,6 @@ panels. No text, no labels, no borders.
   panel stays accurate; the body panels won't match any real clothing the person owns. Still
   useful for identity, less useful if you need clothing consistency.
 
-## Minimal Python example
-
-```python
-import base64
-from openai import OpenAI
-
-PROMPT = (
-    "Convert the person in the photo into a character reference sheet with four panels "
-    "arranged left to right: a close-up of the face, then full-body front, side, and back "
-    "views, on a white background."
-)
-
-client = OpenAI(api_key="YOUR_API_KEY")
-
-with open("input_photo.jpg", "rb") as img_file:
-    resp = client.images.edit(
-        model="gpt-image-2",
-        image=img_file,
-        prompt=PROMPT,
-        size="1536x1024",
-        quality="medium",
-    )
-
-img_bytes = base64.b64decode(resp.data[0].b64_json)
-with open("character_sheet.jpg", "wb") as f:
-    f.write(img_bytes)
-```
-
 ## Using the result
 
 - **For inference** (ComfyUI, `LTX Identity Transfer` node): feed the 4-panel sheet directly as