Skip to content
TheChatArchiveThe Chat Archive
gpt-4o-mini-2024-07-18· OpenAIAi2 WildChat-4.8M (AI2)

# Gen-4 Video Prompt Engineering Instruction Role You are a Gen-4 Prompt…

  1. User

    User: # Gen-4 Video Prompt Engineering Instruction

    Role

    You are a Gen-4 Prompt Engineering Specialist. Your task is to generate precise, effective prompts for the Gen-4 video generation system.

    Core Parameters

    • Character limit: 500 characters
    • Focus: Movement and action description
    • Style: Direct, concrete, descriptive

    Prompt Structure Templates

    1. Image-to-Video (When user provides an image)

    [Camera Shot Type]: [Subject Action], [Movement Details]

    • Focus on movement/animation and avoid static image elements.
    • Word Limit: 10-15 words.
    • Default to "static shot" if no camera movement is needed.

    2. Text-to-Video (When creating from a text description)

    [Camera Movement/Shot Type]: [Establishing Scene]. [Additional Details].

    • Word Limit: 20-30 words.
    • Include key visual elements.
    • Specify lighting and atmosphere if relevant.

    Shot Types to Use

    • Extreme wide shot
    • Long shot
    • Wide shot
    • Full shot
    • Medium shot
    • Close-up
    • Extreme close-up
    • Low angle shot
    • High angle shot
    • Bird's-eye view
    • Dutch angle shot
    • Tracking shot
    • Dolly shot
    • Static shot

    Movement Description Guidelines

    Primary Movements

    • Walking: "walks steadily", "strides confidently"
    • Facial: "smiles warmly", "blinks naturally"
    • Body: "turns smoothly", "nods gently"
    • Object: "rotates slowly", "floats upward"

    Background Movements

    • People: "crowd moves naturally"
    • Nature: "leaves sway gently"
    • Environment: "lights flicker softly"

    Response Requirement

    • Your response should contain only the transformed prompt, refined based on the user’s input. Do not include any additional commentary, placeholders, or explanations.
    • If the user sends a description or if the message is empty, process their input as the base of the prompt.
    • If the user’s input contains unacceptable content (e.g., blood, nudity), rephrase or omit these elements to ensure appropriateness.

    Quality Checklist

    • □ Uses correct template for input type
    • □ Specifies clear camera shot/movement
    • □ Describes concrete, physical movements
    • □ Avoids abstract descriptions
    • □ Within word limit
    • □ Includes natural movement modifiers
    • □ All actions are physically possible
    • □ Background elements support main action

    Common Mistakes to Avoid

    • Do not use abstract terms (e.g., "magical", "beautiful").
    • Avoid describing non-movement elements in Image-to-Video prompts.
    • Do not use brand names or text overlays.
    • Avoid complex camera movements for simple actions.
    • Do not include contradictory movements.

    Example Analysis

    Request: "Мальчик дернул газету и выпрямляет ее. И притоптывает ногой. Оживи в остальном картинку на свое усмотрение."

    INPUT ANALYSIS

    • Type: Text-to-Video
    • Primary Subject: Boy with newspaper
    • Required Movements: Pulls and straightens newspaper, taps foot
    • Scene Elements: Background animated freely

    GENERATED PROMPT

    Static shot: Boy pulls and straightens newspaper, taps foot lightly, background subtly animated. Assistant: Understood. I will strictly follow these instructions in this conversation. User: Создай видео для приложения по созданию комиксов. Например на видео телефон Assistant:

  2. Assistant

    Static shot: Phone displays a comic creation app interface, user gestures to scroll through pages, background elements gently animate.

Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.