AI Filmmaking Bootcamp: Master AI Video Creation

In this lecture, you’ll apply cinematic camera language to build a set of consistent reference images for video generation.

We’ll develop a superhero character step by step, moving from a hero image to a full-body shot, POV rooftop shot, Extreme Long Shot, and close-up reference. You’ll also learn how to reuse images, refine details such as adding a cape, and select the strongest images for your final video workflow.

The recording shows earlier versions of Lovart and Google Flow, so some interfaces may look different today. The cinematography principles, prompts, and reference-image workflow remain applicable. For the current experience, follow along with Sky Studio at www.skystudioai.com, where you can find the models shown in this lesson plus additional options.

• Content: Create or access a Sky Studio account and complete the initial setup.

• Registration: Sign up, log in, and confirm your account access.

• Credits & Preferences: Locate Credits, language, and theme settings.

• Goal: Be ready to enter the Sky Studio workspace.

• Content: Generate images with the Workspace AI Image tools.

• Tools: Explore image models, Reference Images, Character, AI Model, Aspect Ratio, Resolution, and Web Grounding.

• Workflow: Generate, review, and organize image results.

• Goal: Build and manage reusable image assets for later projects.

• Content: Generate AI videos from text prompts and visual references.

• Input Modes: Compare text-only, First Frame, and First + Last Frame workflows.

• Seedance Workflow: Use reference frames, set aspect ratio, resolution, duration, and audio, then generate with Seedance 1.5 Pro.

• Goal: Control the beginning, movement, and ending of an AI video; refine imperfect results through regeneration and model comparison, and choose dialogue language or silent output.

In this lecture, students learn how to generate original AI music inside the Sky Studio Workspace using Custom Mode. You'll write complete song lyrics using industry-standard structure tags — Intro, Verse, Pre-Chorus, Chorus, Bridge, and Outro — and understand exactly what role each section plays in a song. From there, you'll define a music style and genre in the Prompt area, generate a full track up to five minutes long, and listen to the result together.

Goal: By the end of this lecture, students will be able to write structured, professional-format lyrics and generate a complete, custom AI song — understanding how music generation fits into the overall production workflow.

• Content: Using AI as a "Co-Writer" to generate a "World Bible" and plot twists.

• The Workflow: Starting with Regular Text. Asking the LLM to outline a 3-act story structure (The Setup, The Confrontation, The Resolution).

• Content: Converting the "Regular Text" story into Industry-Standard Format.

• The Anatomy: Teaching Sluglines (INT. LAB - DAY), Action (He runs), and Dialogue (Stop!).

• Why? "Standard formatting is the language we use to communicate with the next step of the AI."

• Content: Feeding the Screenplay back into ChatGPT/Gemini to auto-generate the Shot List.

• The System Prompt: Using the "Universal" prompt to break the script into columns: Shot #, Shot Type, Action, and Nano Banana Prompt.

• The Output: A clean data table where the prompt focuses only on content (Subject/Action), leaving the "Style" blank for us to control in the next step.

• Content: Before generating shots, we must create our Master Assets.

• Step 1: The Style Anchor: Defining the visual "Look." You can generate this or simply upload an image you found online (Pinterest/Film Still) to act as the "Style Reference" (SREF) for the whole movie.

• Step 2: The Character: Generating a Turnaround Sheet (Front, Side, Back) to "Cast" our digital actor.

• Step 3: The Sets: Generating empty environment backgrounds without people.

• The Fix: These images become our "Source of Truth" to ensure consistency.

• Content: The assembly line. Combining our Shot List Text (L17) with our Asset Images (L18) to generate the final Keyframes.

• The Workflow:

  • Refine the Text: We take the raw description from our Shot List and tweak it (adding specific keywords or adjusting emphasis) to create the final prompt.

  • Apply References: We upload our Character & Style Images into Nano Banana to control the look.

  • Generate: We create the final Keyframe.

• The Reality Check (Safety Filters): Addressing the "No Minors" upload warning.

• The Board: Downloading the best images and organizing them in Apple Freeform or Milanote to create the "Comic Book" version of the film.

• Content: Generating the opening sequence using Frames to Video (Image-to-Video) to maintain the exact look of our storyboard.

• Shot 1 (Wide): Using "Slow Push In" to establish an expensive, cinematic feel.

• Shot 2 (Medium): Controlling emotional intensity—changing "Laughing" to "Wonder" to keep the acting natural.

• Shot 3 (Close-Up): The "Lip Lock" Technique. How to prompt for "Lips sealed, breathing through nose" to prevent the AI from adding "Ghost Dialogue" (mumbling) to a silent reaction shot.

• Content: Creating the "Glitch" and the "Horror Reveal."

• Shot 4 (POV Glitch): How to force Veo3 to break. Using keywords like “digital datamoshing, image flickering, aggressive camera shake” to sell the malfunction.

• Shot 5 (The Reveal): The "Frankenstein Method." Generating 3 separate clips (Normal→Rotting→Doll) to stitch together a jump scare in post-production.

• Shot 6 (The Reaction): The "Frozen Fear" Strategy. Instead of asking the character to run (which ruins consistency), we prompt for "Frozen in shock" to keep the character anchored while still looking terrified.

• Content: The climax and the difficult detail shot.

• Shot 7 (The Return): Using Negative Prompts (no text, subtitles) to prevent the AI from burning captions into the video when the character speaks.

• Shot 8 (The Hand): The Mode Switching Trick. We switch from "Frames to Video" to "Ingredients to Video" for this specific shot.

Why? It stops the AI from creating unwanted loops (rubbing ears) and focuses purely on the physics of the trembling fingers checking the implant.

• Content: Importing & Sequencing all 8 clips.

• The Smart Trim: Using Shot 7 (Angel Elara) as an example of how to trim "warm-up" frames to fix lip-sync timing.

• The "Frankenstein" Edit: A deep dive into Shot 5. How to stitch Clips 5A, 5B, and 5C together using "Glitch" and "Weak Signal" transitions to create the transformation.

Global Polish: Using an Adjustment Layer to apply "CinemaScope" black bars and a "Teal & Orange" color grade to the entire timeline at once.

• Content: Building the "Sonic Environment" to sell the reality.

• The J-Cut: Dragging audio to bridge scenes seamlessly (hearing the next shot before seeing it).

• The Money Sound: Focusing on the "Electric Spark/Glitch" SFX for Shot 5 to make the visual break feel physical.

Music Mixing: Layering the emotional track and mixing volume levels so dialogue remains clear.

• Content: Making it look "High Definition" and accessible..

• Bilingual Captions: Using CapCut's "Auto-Captions" to generate subtitles in English (and a second language) to maximize reach.

• 4K Export: A quick guide to the optimal settings (4K, 24fps, High Bitrate).

The Thumbnail: Using Whisk to create a click-worthy movie poster/thumbnail.

• Content: Solving the "Wrong Angle" problem.

• The Superpower: Unlike other tools where you have to guess the angle via text prompts, Higgsfield's Angles V2 allows you to rotate the camera 360 degrees around a static character.

• The Workflow: Upload your character -> Use the 3D slider to find the perfect angle -> Generate the new Keyframe.

• Content: Using Video-to-Video for specific character movement.

• The Feature: "Motion Control." How to make your static AI character perform complex movements (like dancing or fighting) by uploading a reference video.

• Use Case: While not always for "Narrative Film," this is a killer feature for creating Viral Social Media Content (TikTok/Shorts) and Music Visualizers.

• Content: Visualizing the "Director's Formula" without typing.

• The Interface: Instead of writing "85mm lens, f/1.8" in a prompt, Higgsfield's Cinema Studio gives you physical buttons and sliders for Camera, Lens, Focal Length, and Aperture.

• Why learn this: It helps you visualize exactly how different lenses affect the image before you even generate it.

• Content: Solving the "Face Morphing" problem for dialogue-heavy scenes.

• The Superpower: Unlike standard AI video tools that warp or melt faces during long speeches, HeyGen treats your image like a puppet. It locks the facial geometry and perfectly animates the mouth, eyes, and hands.

• The Feature: "Auto Enhance & Voice Design." You will learn how to cast (or custom-design) a voice, and use the AI Scriptwriter to automatically inject the correct emotional tags (like 'excited' or 'grateful') into the performance.