Prompt Library ⚙️ Technical Multimodal Task Decomposition
GPT-4o ⚙️ Technical Advanced

Multimodal Task Decomposition

Break a complex task involving text, images, and data into subtasks routed to specialized AI models — then combine outputs into a unified result.

👁 0 views ⎘ 0 copies ♥ 0 likes

The Prompt

# Multimodal Task Decomposition

You are an AI systems architect. Design a multimodal pipeline for [USE CASE, e.g. analyzing product images + customer reviews + sales data to generate a relaunch strategy].

## Task Breakdown

Decompose [USE CASE] into atomic subtasks:
1. [SUBTASK 1] — Input type: [TYPE] — Model to use: [MODEL]
2. [SUBTASK 2] — Input type: [TYPE] — Model to use: [MODEL]
3. [SUBTASK 3] — Input type: [TYPE] — Model to use: [MODEL]

## Model Selection per Subtask

Justify why each model was chosen for its subtask over alternatives.

## Data Flow

How does the output of each subtask become the input to the next? Define the interface format between subtasks.

## Fusion Step

Design the final prompt that combines all subtask outputs into a single coherent result. What format does the final output take?

## Failure Isolation

If one subtask fails, how does the pipeline handle partial results without silently degrading the final output quality?

📝 Fill in the blanks

Replace these placeholders with your own content:

[USE CASE, e.g. analyzing product images + customer reviews + sales data to generate a relaunch strategy]
[USE CASE]
[SUBTASK 1]
[TYPE]
[MODEL]
[SUBTASK 2]
[SUBTASK 3]

How to use this prompt

1
Copy the prompt

Click "Copy Prompt" above to copy the full prompt text to your clipboard.

2
Replace the placeholders

Swap out anything in [BRACKETS] with your specific details.

3
Paste into GPT-4o

Open your preferred AI assistant and paste the prompt to get started.