3.2 Image-to-Video (I2V)

Unlocking High Fidelity: Why starting with an image is the “secret sauce” of professional AI video.

Introduction: The Fidelity Gap

In Lesson 3.1, you learned the language of “Camera Directing” using **Text-to-Video (T2V)** prompts. This is the simplest way to make a video: type words, get motion. But for novice filmmakers, T2V has a major downside: **Loss of Art Direction.**

If you only use a text prompt like “A grizzled lighthouse keeper stands in the rain, dramatic dolly shot,” the video model has to simultaneously invent the character’s face, lighting, atmosphere, *and* motion. This leads to generic, inconsistent visuals. We spent all of Module 1 establishing the **Style Guide** and **Master Character Sheet** so we wouldn’t have to guess. We must use them as our starting point.

That is the purpose of **Image-to-Video (I2V).** I2V allows you to upload an image and use it as the definitive, “Genesis” frame of your video generation. The AI isn’t guessing the look; it is animating **the exact scene you already defined.**

What is I2V?
Image-to-Video is a generation method in which an image serves as the primary structural, compositional, and aesthetic anchor. The text prompt is then used to describe *only* the motion, camera move, or interaction—not the subject itself.

Why I2V is the Standard Workflow

This workflow solves the “awesome” problem by ensuring that high-quality *stills* are turned into high-fidelity *motion.*

Text-to-Video (T2V)

Simple: Just one prompt.
AI defines character, lighting, and style.
High chance of character/style drift.
Generally lower overall visual sharpness.
*Best for generic b-roll.*

Image-to-Video (I2V)

Multi-step: Must first generate the perfect still.
You (The Director) define character, lighting, and composition.
Character/Style consistency is anchored to the input.
Drastically higher facial fidelity and sharpness.
Essential for professional AI filmmaking.

The I2V Workflow

Do not simply generate an image and click “animate.” The professional I2V workflow requires precise inputs at both the image and video generation stages. (Phase 3.2.1)

Create the ‘Genesis Still’ (Image Prompting)

Use the advanced consistency techniques from Lesson 1.3. Take your Master Character Sheet, your Visual Style Guide, and your composition prompt. Iterate in Midjourney or DALL-E until you have the *perfect,* photorealistic, compositionally sharp 16:9 still. This must represent the *start* of the video clip.

Draft the I2V Prompt (Focus on Motion)

Now, write the prompt for the video model (Runway Gen-3, Kling, or Luma). The video prompt is different! **Do not redescribe the subject.** Focus exclusively on describing the camera motion, interaction, or specific movement.

Correct I2V Video Prompt:
The camera performs a slow dolly push-in toward the keeper’s eyes. Minimal subject motion. Volumetric shadows [Style Reference].

Incorrect I2V Video Prompt:[Do not redescribe:] A weathered, grizzled old lighthouse keeper with a scar stands looking at a stormy ocean, [Now the AI is confused]

Combine and Direct (Slider Controls)

Upload your Genesis Still to the I2V input slot. Enter your I2V Motion prompt. Then, use the camera sliders learned in Lesson 3.1 (e.g., set Horizontal Slider to ‘-2’ for a slight pan right). This creates a powerful layer of explicit text instruction and compositional slider control.

🔥🔥 AI Director Pro-Tip: The “Motion Bias Hack” 🔥🔥🔥
Advanced video models allow you to control how much the AI *changes* the input image. If you are animating a face or a close-up, keep **Motion Bias (or Noise Slider)** low (20-40). This forces the AI to strictly preserve the face’s exact details, sacrificing wide motion for fidelity. If you are animating a wide-angle landscape with trees blowing, set Motion Bias higher (60-80) to allow for dramatic environmental recalculation. (Phase 3.2.2)

Lesson Assignment

Your task is to practice the precise transition from still-to-motion. We are not generating video yet. We are locking the “Genesis Stills” and their definitive motion prompts.

Pick **one of your 10 key story beats** (from 1.2).
Use your full Module 1 workflows (CREF character sheet + Visual Style guide) to generate the **perfect photorealistic 16:9 ‘Genesis Still’** for that moment. This image must be compositionally sharp and finalized for production.
Submit your final Genesis Still (.jpg or .png) below.
In the assignment text area, submit the **precise, motion-only I2V video prompt** you will use to animate that specific still.

Module 3: Cinematic Motion