Course Curriculum

Stable Diffusion Deforum: Going Viral on TikTok & Instagram

10sections
53lectures
3h 43mtotal

What you’ll learn

  • Everything important about Deforum and Stable Diffusion

  • Essential background knowledge about Seed, Frames, and more

  • Detailed step-by-step instructions for your first video

  • How to insert yourself into such AI videos

  • How to reshape your own videos using prompts

  • How to use your video as a starting point for animations

  • All parameters like camera axes, zoom, rotations in 2D and 3D, etc.

  • How to create videos that can potentially go viral, and much more

01

Introduction

4 lectures • 8min

  • Welcome!

    0:25

  • Course Overview

    Navigate deforum diffusion basics, from prompts, seeds, and frames to 2d/3d animations, with quick access options (Discord, Google Colab) and a guided path to your first video.

    4:10

  • IMPORTANT: My goal, a tipp and what to expact

    build a community around deforum diffusion, inviting learners to share tips via the Q&A. stay up to date and learn how to make these ai-powered videos for future-ready skills.

    1:40

  • Instructor Introduction: Arnold Oberleiter (Arnie)

    Arnold Oberleitner, aka Arnie, introduces his multilingual background and focus on AI, transformers, diffusion, and automations for small businesses, plus leading German communities and channels with workshops and consulting.

    1:20
02

Background Information: Seeds, Frames, Stable Diffusion, Models and Inspiration

10 lectures • 26min

  • Links you need to know on one place

    0:22

  • What is Stable Diffusion?

    discover how stable diffusion, an open source diffusion model, turns prompts into pictures using Python code. Explore running on GPUs locally or via Google Colab and Deforum extension.

    2:01

  • The technology behind it: The diffusion model

    Explain how the diffusion model learns from images by adding noise and text descriptions, then reconstructs pictures from noise using specific prompts like a gray cat on a beach.

    3:08

  • Why we need Stable Diffusion if we have Dall-E and Midjourney?

    Discover how stable diffusion serves as the backbone for Dall-E, Midjourney, Leonardo AI, and Deforum, and learn to refine prompts from Leonardo and Lexica to create videos.

    1:54

  • What are models: Leonardo.AI as example

    Discover how stable diffusion uses fine-tuned models like the Leonardo diffusion model to create specific styles at 640x832 or 1024x1024 pixels, and learn to find, download, and upload them.

    3:35

  • What is the Seed: Midjourney as example

    Learn how the seed, or starting noise, drives diffusion outputs for precise, repeatable images when the same seed and prompts are used; tweak prompts for subtle variation.

    3:10

  • What are frames?

    Understand that a frame is a single picture of a video, and 30 or 60 fps create motion from still images.

    1:41

  • Prompt Engineering in Stable Diffusion: You need the negative prompt

    Master prompt engineering in stable diffusion by using precise prompts and negative prompts to boost image quality and exclude unwanted elements.

    4:37

  • Finding Inspiration on Lexica and Leonardo AI

    Explore inspiration for prompts on Leonardo AI and Lexica, copy prompts from showcases and likes, and use the seat to remix and refine results with positive and negative prompts.

    2:42

  • The Definition of Learning: Are you a good learner?

    Learning means same circumstances, different behavior; you learn when you change your actions after a cue, not by repeating the same response, and try using ai.

    2:51
03

Acess to Deforum (local and cloud solutions) the discord, and the path setup

6 lectures • 14min

  • Join the discord community

    Join the Deforum community on Discord to stay up to date, access tutorials, official links, a quick user guide, and chat with other users.

    0:52

  • Local Installation of Stable Difusion with Python, Github and Autmatic 1111

    Learn to install Stable Diffusion locally with Python, Git for Windows, and automatic 1111, including models 1.5 and 2.0, while also exploring cloud-based options for GPU-light rendering.

    1:55

  • RunDiffusion: A cloud solution that works

    Explore run diffusion as a cloud solution to render videos with rented GPUs and per-hour pricing. Learn why Google Colab offers a cheaper, fast connection for rendering deforum projects.

    1:45

  • Colab Notebook and Google Drive

    Learn to set up Deforum on Google Colab using Google Drive to run notebooks, install models, and generate prompts and outputs with Stable Diffusion 0.7.

    5:38

  • The pricing in Colab Notebook

    Explore Colab notebook pricing for stable diffusion deforum, detailing Tesla T4 GPU access, Colab Pro options, and hours billed at two hours per hour for rendering.

    2:32

  • Path Setups: Models and outputs

    Set up your diffusion workflow by configuring the model path and output path to organize ai models and generated outputs; learn to upload and use custom model checkpoints.

    1:46
04

How does deforum work? Step by step to your first video

16 lectures • 1hr 1min

  • How deforum works, what does deforum?

    Deforum diffusion extends stable diffusion to generate customizable video frames. Start with a frame or your image, set seeds, camera and rotations, render, and tweak parameters to create video sequences.

    2:11

  • Deforum overview and quickguide

    Explore the deforum overview with 2d and 3d motion settings, including zoom, translations, perspective flip, and camera rotations, guided by a quick overview and practical examples.

    2:52

  • Basics of animations: Angle and zoom as example

    Master the basics of deforum animations by configuring angle and zoom across frames, using 2D inputs, frame ranges, and motion parameters to create clockwise and counterclockwise rotations and static segments.

    4:11

  • Kamera movements in 2D and 3D

    Adjust translation x and y to move the camera, and use translation z for zoom; explore 2d rotations, while 3d rotations are not used yet, with values -8 to 8.

    3:22

  • What does the formula in the bracket mean?

    Explore how the bracket formula drives camera movement and wave-like animations in Deforum, using x axis waves, frame ranges, and keystring generators to craft dynamic sequences.

    4:18

  • Noise schedule and strength schedule

    Adjust noise and strength to control frame similarity and zoom speed in stable diffusion animations. Use a 0.65 strength with around 0.02–0.03 noise as a starting point.

    4:51

  • Coherence/Cadence: Make the videos smoother

    Explore coherence and diffusion cadence to smooth videos. Adjust color coherence with HSV or Lab, and set cadence (1 or 2 diffused frames) to balance render time and output quality.

    2:17

  • The prompt cell and negativ prompts

    Use the prompting cell to input prompts and animation prompts, enabling frame-based transitions in stable diffusion, with two-dimensional, three-dimensional or video inputs and negative prompts to exclude unwanted features.

    5:30

  • UPDATE: v0.7.1 brings a new Prompt Box

    Upgrade to version 0.7.1 of the Deforum prompt box, simplifying prompts and negative prompts, removing animation prompts, and preserving batch settings for 2D, 3D, or video inputs.

    1:41

  • Load and image settings & the resolution

    Set load and image settings to the model's 512 by 512 resolution, or 576 by 1024 for Instagram and TikTok; keep bit depth at eight to save render time.

    2:58

  • Sempling settings, the seed and steps

    Set a specific seed to track generations, avoiding the minus one random seed. Use the default sampler and 40–70 steps, with a scale of 5–15 for balanced prompts.

    3:10

  • Batch settings and prompt weigths

    Master prompt weights to emphasize specific words and control strength in prompts, then configure batch settings to set variations and seed behavior (iter, fixed, random) and where images save.

    3:41

  • The last settings befor we render the video

    Explore the final render settings, including init, mask options, and default exposure, then configure frames per second, save paths, and max frames to prepare your video render.

    3:26

  • We render the first video

    Render the first video by configuring 200 frames with rotation, zoom, and camera translation, export frames to Google Drive, and assemble a 12 fps animation.

    6:17

  • Letting people age and the resume animation

    Learn to age a subject in deforum diffusion by creating a 100-frame 2d animation from a young woman to older ages, adjusting prompts, strength, and frames with resume animation.

    9:24

  • Your learning experience

    Engage your learning experience in Stable Diffusion Deforum by using the forum, create your own animation, describe prompts, start with a static picture, set zero camera movement, then adjust strength.

    0:40
05

Repurpose your own photos and videos to start your animations

3 lectures • 26min

  • Init Image: Start with your own foto as keyframe

    Start your animation with your own photo by uploading it to Google Drive and linking it in Google Colab, using an init image and prompts morphing from beard to zombie.

    9:51

  • Start with your own videos as starting keyframe

    Start with your video as the starting keyframe, export last frame, use Google Colab to animate in 2D with prompts Greek god, zombie, venom, and finish in Shotcut with music.

    6:53

  • Video Input: Lay prompts over videos

    Overlay prompts onto a video in Google Colab by loading it and extracting frames. Apply frame prompts and adjust strength and frame rate for cyberpunk or Walking Dead styles, render.

    9:14
06

Advanced Feauters: 3D Animations, New Models and ChatGPT as assistent

3 lectures • 21min

  • The first 3D video animation

    Master 3d video animation using stable diffusion deforum, with keyframes, camera moves, and frame-by-frame prompts to morph from a hoodie figure to Ghost Rider and zombie.

    8:52

  • ChatGPT as personal assistent to write animations

    Leverage ChatGPT as a personal assistant to craft 3D animations in Deforum, guiding diffusion prompts, camera angles, and 3D rotations for engaging videos.

    8:07

  • Models: Use other models for specific outputs

    Learn to use your own models in stable diffusion with deforum. Upload models to Google Drive, choose custom in Colab, and apply prompts and negative prompts.

    3:43
07

Keyframe string generators for faster Workflow and Music animations

2 lectures • 24min

  • Keyframe string generator: Write your animations faster

    Learn to use a keyframe string generator to create frame-by-frame animations across translation, zoom, and 3d rotations, with prompts and control-point tweaks.

    11:45

  • Audio keyframe string generator: Synchronize music and animation

    Synchronize music and animation in stable diffusion deforum using an audio keyframe string generator to map drum hits to zoom and align frames per second with prompts.

    12:08
08

Complete project and what you should create

3 lectures • 21min

  • Complete projekt

    Learn to build a full 3d animation project with Stable Diffusion Deforum in Google Colab, driving prompts, zoom, translation, and rotations across frames to create a viral promo video.

    14:52

  • What you shold create and where you can get inspiration

    Explore what you can create with forum diffusion: 2D and 3D animations, overlay prompts on your videos or stock videos, and camera rotations, then seek inspiration on TikTok and Instagram.

    3:02

  • Learning experience in 3D

    Create a 3d video using your own or provided models, animate with keyframe string generators, add music, and publish on Instagram or TikTok with platform-appropriate resolutions.

    2:43
09

Alternatives to Deforum

3 lectures • 21min

  • Kaiber: A solied and easy alternative

    Discover Kyber as a fast, affordable alternative to warm diffusion, letting you upload a portrait, enter prompts, and overlay results to create simple 2D videos.

    3:16

  • Loopsie and Disco or Wrapfusion

    Explore diffusion-based animation tools, from Loopsy’s mobile option with limited control to disco diffusion and RGB diffusion for overlaying prompts on videos, with deforum diffusion offering superior 3d rotations.

    1:55

  • UPDATE: Stable WrapFusion Tutorial [In Google Colab]

    Learn to apply stable wrap fusion in Google Colab by using a Patreon-sourced Colab notebook, extract and transform video frames with prompts, and render results in Blender or Shotcut.

    15:49
10

Whats next and my personal thank you

3 lectures • 2min

  • What is next: Create and share your animations

    Explore the forum, create animations with Deforum diffusion, and share them on TikTok, Instagram, or YouTube to attract followers and grow your audience.

    0:57

  • Thank YOU!

    Thank you for watching this course; ask questions anytime as I answer quickly, and go explore yourself as AI shapes the future, placing you at the forefront.

    0:43

  • Bonus

    0:30