Resources

Last Updated April 24th, 2025

Variance’s Current Workflow:

I. Pre-Production — Planning the Vision

1. Story & Script

  • ✍️ Write a story concept or screenplay

  • Tools: Final Draft, ChatGPT, Google Docs, Scrivener

2. Storyboarding & Shot List

  • 🖼️ Create AI stills or rough thumbnails to plan shots

  • Tools: FLUX, Midjourney, Frames, Mystic 2.5, Sora (image gen)

3. Visual Planning

  • 🎨 Generate reference images for key visuals and style

  • Tools: FLUX, Sora, Frames, Midjourney, Photoshop

4. Prompt Development

  • 🧠 Write detailed video prompts (scene, motion, lighting, mood)

  • Use prompt guides for: Ray2, Sora, Hailuo, Kling, Hunyuan, WAN2.1, Veo2, Runway Gen4 with ChatGPT to help with this.

II. Production — AI Video Creation

1. Scene Generation (Text-to-Video or Image-to-Video)

  • Tools:

  •   - 🎥 Higgsfield AI: DOP Level Control, cinematic

  •   - 🎥 Ray2: Realistic motion, cinematic angles

  •   - 🎥 Sora: Story-driven multi-scene realism

  •   - 🎥 Hailuo (Director Mode): Advanced camera control

  •   - 🎥 Veo2: Cinematic film language, long scenes

  •   - 🎥 Runway Gen-4: Fast, cinematic AI video

  •   - 🎥 WAN2.1: Open-source, text/image-to-video

  •   - 🎥 Kling 2.0 & Hunyuan: Fast and stylized visuals

2. Chaining & Extending Scenes

  • 🔄 Create continuity with Extend tools or Remix/Storyboard mode

  • Tools: Luma Ray2’s Extend, Sora’s Remix, Veo2’s sequence chaining

3. Iteration & Revisions

  • 🔁 Refine prompts based on test results

  • Update scene design or prompts as needed

III. Post-Production — Editing & Delivery

1. Video Editing & Assembly

  • 🧩 Stitch AI-generated clips into narrative flow

  • Tools: Premiere Pro, DaVinci Resolve, CapCut

2. Sound Design & Music

  • 🎵 Add AI music, sound effects, or voiceover

  • Tools: Suno.ai, ElevenLabs, Epidemic Sound

3. Color Grading & FX

  • 🎛️ Add LUTs, transitions, overlays, polish shots

  • Tools: After Effects, Runway, DaVinci

4. Final Export & Delivery

  • 📤 Render final version for YouTube, festivals, or clients

  • Add SynthID (if required for Veo2 or ethical use)

Best Large-Language Models:

OpenAI’s ChatGPT

Anthropic’s Claude

Google’s Gemini

DeepSeek (open-source)

X-ai’s Grok

All-In-One for LLMs:

Perplexity

(Has Access To Some Of The Above Models At One Price)

Prompts For Getting GPT To Help You Learn / Have Better Conversations

Video Models

S-Tier:

RunwayML’s Gen-4 (yay!)

Google’s Veo2 [via Freepik & FAL & Krea]

Kling 2.0 [elements!]

Higgsfield AI

A-Tier:

Alibaba WAN 2.1 [open-source, LoRA’s]

Tencent Hunyuan [open-source, LoRA’s]

Luma Ray2

Hailuo-01 Director

Pika 2.2

OpenAI’s Sora

RunwayML Gen3 [for act-one and extend]

All-In-Ones:

Freepik

Krea

FAL (more complex, but more options)

Other Models:

Leonardo.Ai (Photos, Videos, Realtime Canvas)

LTXStudio (Good all-in-one for shorts)

InVideo  (Turn any idea or content to video, instantly with AI)

Upscalers

Magnific (Image Upscale / Structure Reference)

TopazLabs (Video / Image Upscale Software)

Krea.ai (Photo / Video Upscaling)

Audio

Suno AI (Create Full Songs with Lyrics or Instrumentals)

Udio AI (Create Full Songs with Lyrics or Instrumentals)

ElevenLabs (Dubbing, Voice Cloning, Custom Voices)

Graphics / Images

Krea (Flux, Train Your Own Models)

Magnific Mystic 2.5 (Images, Style Transfer)

Freepik (Flux, Train Your Own Models)

Leonardo.Ai (Photos, Upscaling, Flux)

Midjourney (Images, Edit, Consistency, moodboards)

Canva (Thumbnails, graphics)

Writing

SudaWrite (Writing Assistant)

Any of the Top LLM’s Above [especially GPT 4o]

Etc / Miscellaneous / Advanced

CivitAI (Download Open Source Models & Workflows)

YouTube Specific


Vizard (Shorts from Long Form Content)

VidIQ (Titles, Thumbnails, Ideas, Stats)

CapCut (Free Editing Software)

Davinci Resolve (Free Editing Software)

HeyGen (Dubbing to Different Languages, Avatars)

(note, some of these are affiliate links and will benefit our channel! Thank you!)

I will continue to add more tools for you to use to speed up your process.

My current spend per month for all of this crap: $794.99!

GenAI Quick Reference: Essential Terms

  1. Generative AI (GenAI):
    AI technology that creates new content (images, text, video, audio) from scratch based on patterns learned from data.

  2. Model:
    The AI system trained on large datasets to understand patterns and generate outputs. Examples include Stable Diffusion, Midjourney, DALL·E, ChatGPT, etc.

  3. Checkpoint (Model Weights):
    A snapshot of a trained AI model at a certain point. Different checkpoints can drastically change the style or quality of AI-generated content.

  4. Prompt:
    A text input given to a generative AI model to instruct or guide what it should create. Effective prompts typically combine clear instructions, style descriptions, and context.

  5. Seed:
    A numeric value that sets the starting point for random generation. Using the same seed with identical settings ensures consistent, reproducible outputs.

  6. LoRA (Low-Rank Adaptation):
    A small, specialized add-on trained on top of a base model to fine-tune it for specific styles, people, objects, or concepts. LoRAs allow quick customization without retraining an entire large model.

  7. Embedding (Textual Inversion):
    A special token trained to represent a concept or style within a model. For instance, you can create an embedding named "NovaStyle" to consistently generate content with your signature visual style.

  8. Sampling Steps:
    The number of iterative steps the AI takes to create an image. Higher sampling steps often yield clearer and more detailed outputs, but take longer to generate.

  9. CFG Scale (Classifier-Free Guidance):
    Controls how strictly the AI follows your prompt. Lower values give the model more creative freedom (potentially less accurate), while higher values strictly follow the prompt (potentially overly rigid).

  10. Inpainting:
    Using AI to edit or replace specific areas of an existing image.

  11. Outpainting:
    Expanding an existing image by generating additional context around its edges.

  12. Diffusion Models:
    Generative models (like Stable Diffusion) that create images by iteratively refining a noisy starting point into a clear, detailed output.

  13. GAN (Generative Adversarial Network):
    A model composed of two neural networks (a generator and a discriminator) that compete against each other to produce realistic outputs.

  14. Fine-Tuning:
    Adjusting an existing AI model to perform better on a particular task or adapt to a specific style by training on new, targeted data.

  15. Open-Source:

    A philosophy or practice where software, projects, or creative works make their original source materials publicly available. Anyone can freely view, modify, and distribute these resources, promoting collaboration, transparency, and community-driven improvement.