Replicate: Generating LoRa YouTube Thumbnails

Description

**This tutorial will show you how to use the Trained LoRa Model to generate YouTube Thumbnails. 

Part 1: Training The LoRa
https://youtu.be/QSgC7Db84ZM

Replicate:
https://replicate.com

**Fast-flux-trainer:
**
https://replicate.com/replicate/fast-flux-trainer/train

**LoRa:
**
Low Rank Adaptation

———————

Gear I Use:

https://mikemurphy.co/resources


To try or buy Adobe After Effects CC 2025:
https://mikemurphy.co/adobe (affiliate link)

Terrapin Textures:
https://terrapintextures.com

⭕️ Check out my Domestika Course on Adobe Audition:
https://mikemurphy.co/domestika

——————

✅ Chapters:
00:00 Intro: About The Tutorial
00:16 Review of Part 1
00:49 Benefit of Training LoRa
01:37 Generating YouTube Thumbnails
02:41 Prompting Using Trigger Words
03:11 Change Aspect Ratio
03:45 Tweak Prompts
04:21 Download & Add To Examples
04:45 Hallucinations + LoRa
05:46 Tweaking The Cowboy
06:34 Podcast Studio Example

How To Train A LoRa

1. Go to https://replicate.com
2. Sign in with GitHub account
3. Go to ‘Dashboard’
4. Click on ‘Trainings’
5. Click on Fast-Flux Trainer
6. Destination: Create New Model (for example…aihandyman-lora-v1)
7. Upload .zip folder of images
8. Create Trigger_Word (i.e. mkupy)
9. Click ‘Create Training’
10. Click on Trained Model
11. Enter Trigger Word + Prompt
12. Cmd + Enter
13. Have fun!