How to Send bulk prompts automatically to ChatGPT?

[User7]

i want to have ready prompts like

create image of a man walking in park

create image of showing him not sitting on chair

create image of him doing exercise

etc..

it has to be in one window as all images are continue images
i want an extension or something that sends the prompts from a list of prompts with delay of getting answer + auto save image will be good

anyone know something like this?

 

[HotMilfNearYou]

Probably look this way to find your own: n8n, Make, Zapier, or other tools like that for triggering and syncing actions into your workflow, along with an OpenAI API key, and likely something like Google Sheets to store the prompts. Then the workflow starts by taking the first prompt, sending it to GPT, where GPT creates the image, and the image is saved in your cloud or something similar. It will be too complex and lengthy for me to explain in detail, but you can do this:


1: Go on ChatGPT and instruct it to act like a prompt engineer. Explain that you need it to craft a prompt for a new chat. In the new chat, ChatGPT will assist you step by step in figuring out how to accomplish what you asked.


You can literally say something like, "Here is my situation: [explain what you asked here and paste your message and my response]."


Demand for the prompt in the next chat will have an expert role specifically assigned for the task (very important; that's why you need another chat).


Be explicit about the fact that the actual task is creating a prompt for the next chat, where the ultimate task will be completed, or sometimes, AI will try to be "helpful" by running the final task immediately.


Tips:
-Use the exact same model for both chats, with the web search option activated.

-Tell the prompt engineer that you want the “expert” in the next chat to find relevant YouTube videos, when possible, to illustrate each part it explains to you. Specify that the videos need to illustrate the principle but do not necessarily have to use the same exact tools you will. Additionally, ask for an explanation of the differences. If you encounter difficulties regarding this part, you can try using Gemini to see if it finds better videos.

- Conclude the initial prompt to the prompt engineer by instructing it to ask you questions to clarify the situation and to never start answering you without asking the questions. If it bypasses questions, just tell it that it failed because it needs to ask questions.


2: It will give you a list of very precise questions, sometimes more than ten, so it understands your skill level and exact situation. You answer those questions, and it will provide you with a world-class level prompt for the next chat.


3: Open a new chat, paste the prompt from the first chat output, and let it lead you step by step. Ask questions when you need. Don’t believe everything; if something it tells you doesn’t work, search directly with DuckDuckGo or Google for what you want. You will usually find solutions and/or reasons for the problem on forums, blogs, or Reddit. Just explain what doesn’t work and share what you find.

I've been learning many tech concepts like this since 2023.


Edit: typos corrected by chatgpt, then final edit manually by me

 

[tubeboost]

I think the best idea is to use API.

 

[User7]

I think the best idea is to use API.

API is treating each image as separate command and do not know the history of the image creation so it cannot work

It is a simple way - it can work with recording mouse clicks on a browser but im looking for something simpler like chrome extension...?

 

[theusualkeysersoze]

N8N is perhaps fit for purpose, but you might want to use manus as base line, I can confidently say it is very decent, if you know what you are doing.

If you need this as app or ran locally, do not understimate what this takes to get it right, AI or not AI.

There is a lot of logic handling to be done and you want to have a solid architecture for this. Error handling, concurrency etcetc

A paid plan for some AI that can generate these fast and a layer also paid to process the logic, else this will not go anywhere.

People say AI and vibe coding is not going anywhere, as a senior full stack lead dev, I can tell you they absolutelly do, in the hands of a senior dev, there is not one doubt, it gets a but funky on larger teams etc, but the paid for interfaces are very, very decent already, my point is, you need to know how to do this structurally even without AI, the programmatic logic and API choices, you need to be able to handhold the AI and the non paid models have no memory, no context and very hard limits when it comes to image generation.

 

[User7]

is there a chrome extension that can do this?

 

[Proxy Cake]

Probably easiest with an automation extension like Automa to paste prompts from a list and wait for the image to generate, then trigger a save each cycle.