Complete inventory of 40+ AI video tools, ready-to-use workflows, legal guidelines, and terminology reference.
All 40+ tools mentioned in the masterclass, organized by category with pricing and capabilities.
Capabilities: Up to 20 seconds, native audio, photorealistic quality, 1080p output
Best For: High-quality short clips, establishing shots, cinematic scenes
Limitations: 20s max, requires ChatGPT Plus subscription
Visit OpenAI Sora →Capabilities: 8 seconds with sound, text-to-video, excellent quality
Best For: Quick clips with native audio, reaction shots
UK Note: Image-to-video NOT available in UK, text-to-video works
Visit Google Veo →Capabilities: 16 seconds, native audio, best physics simulation
Best For: Future use - currently not publicly available
Limitations: No public access yet, research preview only
Learn More →Capabilities: 10 seconds, cinematic control, motion brushes, camera controls
Best For: Professional creators, fine control over camera movement
Strengths: Best-in-class camera controls, consistent style
Visit Runway →Capabilities: 8 seconds, sound effects, lip-sync, scene explosion effects
Best For: Quick tests, stylized content, adding effects
Strengths: Free tier generous, easy interface, fast generation
Visit Pika →Capabilities: Up to 2 minutes! Best for longer sequences
Best For: Extended scenes, full conversations, narrative sequences
Strengths: Duration is unmatched, good character consistency
Visit Kling →Capabilities: 5 seconds, very fast generation (120 seconds)
Best For: Rapid prototyping, testing many variations
Strengths: Speed, good free tier, smooth motion
Visit Luma →Capabilities: 4-6 seconds, unlimited generations on paid plan
Best For: High volume creators, testing many prompts
Strengths: Unlimited makes it cost-effective for bulk work
Visit Haiper →Capabilities: Access to 5+ models in one subscription
Best For: Comparing different models, budget-conscious
Strengths: Multiple engines, cheap access to variety
Visit Freepik →Capabilities: Most realistic avatars, 300+ voices, custom backgrounds
Best For: Historical characters, talking heads, professional videos
Free Trial: 1 minute video credit
Visit HeyGen →Capabilities: Animate photos, good lip-sync, 120+ voices
Best For: Animating historical portraits, budget projects
Strengths: Cheapest avatar option, good quality
Visit D-ID →Capabilities: 200+ avatars, templates, team features
Best For: Professional/corporate content, templates
Strengths: Most polished, best for business use
Visit Synthesia →Capabilities: AI script writer, conversation mode, 70+ languages
Best For: Training videos, multi-lingual content
Strengths: Built-in scriptwriting, conversation features
Visit Colossyan →Capabilities: Custom avatars, PPT to video, article to video
Best For: Repurposing content, custom avatars
Strengths: Content conversion tools, API access
Visit Elai →Capabilities: Voice cloning, 29 languages, ultra-realistic
Best For: Custom character voices, voice cloning
Free Tier: 10,000 characters/month (10 mins)
Visit ElevenLabs →Capabilities: 900+ voices, voice cloning, commercial rights
Best For: Podcast creators, audiobooks
Strengths: Huge voice library, clear licensing
Visit Play.ht →Capabilities: 120+ voices, video sync, team collaboration
Best For: Teams, professional narration
Strengths: Good team features, voice changer
Visit Murf →Capabilities: 13B parameters, text-to-video, image-to-video
Hardware: 48GB+ VRAM (dual GPUs or cloud)
Fine-Tune: SkyReels V1 specializes in humans - perfect for historical characters!
HuggingFace →Capabilities: 10B parameters, 5.4s at 30fps, Apache 2.0 license
Hardware: 24GB+ VRAM (single RTX 4090)
Strengths: Most accessible high-quality model
Download →Capabilities: Optimized for speed, 24fps, image/video-to-video
Hardware: 12GB VRAM minimum (RTX 3060)
Strengths: Runs on consumer hardware, very fast
Download →Capabilities: Multiple variants (1.3B-7B), excellent i2v
Hardware: 8GB+ VRAM (small model)
Strengths: Very efficient, good for limited hardware
Download →Capabilities: 11B params, Sora-like architecture
Hardware: 40GB+ VRAM
Strengths: Academic research, experimental features
GitHub →Capabilities: Auto-captions, effects, transitions, templates
Best For: Beginners, social media content, quick edits
Strengths: 100% free, easy interface, AI features
Download CapCut →Capabilities: Edit video by editing text, AI voices, studio sound
Best For: Podcasters, creators who think in text
Strengths: Unique transcript editing, voice cloning
Visit Descript →Capabilities: Hollywood-grade editing, color grading, effects
Best For: Serious creators, professional quality
Strengths: Free full version, industry standard
Download →Capabilities: Upscale to 8K, denoise, deinterlace, frame interpolation
Best For: Enhancing AI-generated or old footage
Strengths: Best upscaling available, one-time purchase
Visit Topaz →GPUs: RTX 4090, A6000, A100
Best For: Beginners, on-demand usage
Strengths: Simple interface, community templates
Visit RunPod →GPUs: Marketplace - various options
Best For: Budget-conscious, flexible needs
Strengths: Lowest prices, many GPU choices
Visit Vast.ai →GPUs: A100, H100
Best For: Heavy models, professional projects
Strengths: Most reliable, fastest GPUs
Visit Lambda →Proven workflows for different video types and historical periods.
Best for: Comedy routines, historical commentary
Write 150-word monologue in character's voice
Select period-appropriate voice, generate audio
Upload portrait, sync with voice, generate video
Add subtitles, music, export
Best for: 3-4 minute episodes with story arcs
4 scenes × 45-60 seconds each. Outline full story
Create 4 avatar clips in HeyGen. Batch process to save time
Use Sora/Kling for establishing shots, transitions
Stitch in CapCut, add transitions, music, subtitles
Best for: Debates, interviews between historical figures
Back-and-forth script between 2 characters
Generate each character separately in HeyGen
Edit conversation by cutting between speakers
Optional: Show both characters simultaneously
Best for: High volume, cost-sensitive projects
Deploy Mochi 1 or HunyuanVideo on cloud GPU
Create 5-10 clips in one session (saves money)
Save all files locally, end GPU rental immediately
Use DaVinci Resolve (free) for final assembly
Legal requirements and ethical guidelines for AI-generated content in the UK.
UK Law Requirements:
"This video features AI-generated characters and voices. Historical figures are portrayed for entertainment purposes."
Portrait Rights:
YouTube:
TikTok:
Instagram:
DON'T:
DO:
Tool Licensing:
Always read the Terms of Service before monetizing content!
Plain English explanations of AI video terminology.
A digital character (usually a talking head) generated by AI. In this masterclass, avatars are historical figures brought to life from portraits.
Supplementary footage that plays while narration continues. Examples: establishing shots of castles, close-ups of objects, transition scenes.
Creating multiple videos in one session. More efficient than generating one at a time. Saves money on cloud GPU rentals.
A technique for guiding AI generation with reference images. Helps maintain consistency across multiple clips.
AI-generated video where a person's face is replaced with another. Requires consent if using real people. Not used in this masterclass.
The AI technology behind most video generators. Works by gradually removing noise from random pixels until a coherent video emerges.
Training an AI model on specific data to specialize it. Example: SkyReels is a fine-tune of HunyuanVideo specialized for human characters.
Frames per second. Higher = smoother motion. Most AI videos are 24-30fps. Some tools offer 60fps for ultra-smooth results.
Filming subject on solid green background, then replacing that background digitally. HeyGen can generate avatars with green backgrounds.
AI that animates a still image into a video. Used to bring historical portraits to life.
Running the AI model to generate output. "Inference time" = how long generation takes. Faster inference = less cloud GPU cost.
The mathematical "imagination space" where AI models create content. Not important for users, but you'll see this term in technical docs.
Matching mouth movements to audio. HeyGen and D-ID do this automatically. Quality varies - HeyGen is best.
A lightweight fine-tuning method. Adds specialized capabilities to a model without retraining the entire thing. Much faster and cheaper.
The "size" of an AI model, measured in billions (B). More parameters generally = better quality but needs more VRAM. Example: Mochi 1 has 10B parameters.
The text description you give the AI. Good prompts = better results. Example: "King Henry VIII sits on throne, looking confused at smartphone"
The process of computing the final video file. Can take 2-10 minutes depending on length and quality. Also called "export."
Video dimensions in pixels. Common sizes: 1080p (1920×1080), 720p (1280×720), 4K (3840×2160). Higher = better quality but larger files.
A number that controls randomness in AI generation. Same prompt + same seed = identical output. Useful for consistency.
Content created or modified by AI. Includes AI-generated videos, voices, and images. Must be labeled as such in UK.
AI that converts written text into spoken audio. ElevenLabs is the best TTS for character voices.
AI that creates video from text descriptions. Sora 2, Veo 3, Runway Gen-3 are all text-to-video models.
Unit of text processing in AI. Roughly 0.75 words = 1 token. Some tools charge by tokens instead of time.
When AI-generated humans look almost-but-not-quite real, causing discomfort. Good lip-sync and natural movement help avoid this.
Increasing video resolution using AI. Topaz Video AI is the best tool for this. Can turn 480p AI output into 1080p or 4K.
AI that transforms existing video into a different style. Example: making a real video look like a painting.
Training AI to replicate a specific voice. ElevenLabs can clone your voice from 1 minute of audio. Requires consent if cloning others.
Video RAM - memory on your GPU. More VRAM = can run larger AI models. Open source models need 12-48GB+ VRAM.
The trained data of an AI model. "Downloading the weights" means getting the model files. Large files - often 10-50GB.