• print stickers

• image generation API

  or

  imagegenAPI

  ;

  API

  is case-insensitive, and spaces between "image generation" and "API" are allowed.

Please add the customer service WeChat with the note "heart-sticker", and the assistant will solve your problem promptly: 77213305

• Enter the text branch when the user explicitly mentions "text images", "text stickers", "text posters", "font images", or "text effects".
• Enter the image branch when the user uploads or references photos, portraits, pets, or object images and requests style conversion.
• If the intention is unclear but an image is provided, default to the image branch; if the intention is unclear and only text is provided, first confirm with one sentence whether the user wants to create text stickers.

1. Check the input image. If no image is provided, only ask the user to upload an image and pause; do not generate a substitute image out of thin air.
2. If the image comes from a local path, use

   view_image

   to view it first. If the image is from the current conversation, observe it directly. Do not claim to recognize details that cannot be seen.
3. Read references/image-styles.md. Recommend 3–5 most matching styles based on the subject, background complexity, and usage, using the format "Number + Style Name + One-sentence Effect Description". Add "Custom Style" and "Remove Background" if needed.
4. Ask the user to select a number or style name. Even if a certain style is obviously the most suitable, you must wait for confirmation.
5. After the user confirms, extract the corresponding original prompt from the reference file and organize it into the following generation instructions:

   • Use case: style-transfer

     ; use

     background-extraction

     for pure background removal.
   • Clearly state whether the input image is the editing target or reference image.
   • Take the original prompt as the core task without changing its style semantics.
   • Add subject retention constraints: retain visible identity features, number of subjects, posture, clothing, and key accessories; only remove or retain the background if explicitly required by the original prompt.
   • Avoid adding new people, repeated limbs, wrong facial features, garbled characters, watermarks, and irrelevant text. Do not delete bubble or text elements without permission if the original prompt explicitly requires them.
6. Call Codex's built-in

   image_gen

   :
   • If all target images have local paths, pass

     referenced_image_paths

     .
   • If the target image only exists in the current conversation, pass the minimum

     num_last_images_to_include

     that can cover the target image.
   • Do not pass both at the same time.
7. Display the result, and briefly report the style name and final prompt. If the user requests fine-tuning, only modify one clear variable and generate again.

1. Collect the text that must be presented verbatim. If the user does not provide specific text, first ask for the text content and pause.
2. Read references/text-styles.md. Recommend 3–5 styles based on the number of words, tone, and usage, using the format "Number + Style Name + One-sentence Effect Description".
3. Ask the user to select a number or style name; generation can only proceed after confirmation.
4. Replace all

   {text}

   in the reference file with the user's original text without polishing, modifying words, or adding punctuation. For custom styles, combine the original text with the user's description into a clear task.
5. Form a brief generation instruction:

   • Use case: logo-brand

   • Text (verbatim): "<User's Original Text>"

   • Original style prompt
   • Require the text to be clear, complete, and in correct order; no extra text, watermarks, or English placeholders are allowed.
6. Call Codex's built-in

   image_gen

   to generate a new image; do not pass a reference image unless the user explicitly uploads a layout or style reference.
7. Display the result and review the text. If typos, missing words, or incorrect word order are found, make a correction only for text accuracy using the same style.

• Portrait photos: Prioritize cute stickers, Pixar 3D, Ghibli, and illustrations.
• Pets: Prioritize cartoon stickers, cute stickers, clay style, and pixel art.
• Dynamic or youthful trends: Prioritize neon future, graffiti, and retro comics.
• If the user wants to retain the original scene: Prioritize neon future, graffiti, and children's drawings; clearly remind that these styles retain the background.
• If the user wants a clean, croppable result: Prioritize cartoon stickers, cute stickers, Pixar 3D, and illustrations; "Remove Background" can also be listed separately.
• Short slogans: Prioritize trendy vitality, classic calligraphy, artistic graffiti, and cute comics.
• Long sentences: Prioritize text posters, classic calligraphy, and childlike casual style; remind that longer text is more prone to typos.
• Warm and healing phrases: Prioritize meme text, cute comics, and childlike casual style.
• Do not display the complete prompt library at once; only expand it when the user requests to view all or audit the source.

• When the user requests to view effects, compare all styles, or create a style selection card, read references/case-gallery.md.
• Cases are only used to illustrate the visual differences of each style; do not copy case characters, actions, compositions, or text into new tasks.
• When generating new images, still rely on the user's materials and current needs; cases cannot replace the reference images uploaded by the user.

• Do not call the Jiemeng interface built into the heart-sticker repository, nor do you need its account, points, or secret key; only reuse prompts and call Codex's built-in image generation.
• Do not count "Remove Background" as a repository prompt; it is an independent image editing capability.
• Follow the image saving rules of the current project. Display directly only for preview; project materials need to be saved to the project-specified directory without overwriting existing files.
• For sources, style IDs, and extraction ranges, see references/source-manifest.md.