Difference between revisions of "Fal.ai"
(→Flux LoRA Portrait Trainer) |
|||
| Line 41: | Line 41: | ||
===Flux LoRA Portrait Trainer=== | ===Flux LoRA Portrait Trainer=== | ||
| − | The [https://fal.ai/models/fal-ai/flux-lora-portrait-trainer flux lora portrait trainer] is used to train the model by uploading images of a person, animal or object. After this you may use [https://fal.ai/models/fal-ai/flux-lora flux lora] to generate new images. | + | The [https://fal.ai/models/fal-ai/flux-lora-portrait-trainer flux lora portrait trainer] is used to train the model by uploading images of a person, animal or object. After this you may use [https://fal.ai/models/fal-ai/flux-lora flux lora] to generate new images (may be you would like to increase 'Num Inference Steps' from 28 to 40, if you don't like the result you may increase the 'Scale' form 1 to 1.2). |
Revision as of 19:13, 3 October 2025
Contents
What is fal.ai
fal.ai is a pay per use (no subscriptions) AI.
Text to Image
Minimax-Image
The minimax-image is a cheap (but slow) hyper realistic (but not stylised) image generator.
Flux
The Flux is a very popular image generator by Flux AI. Which is faster (but more expensive) and able to generate realistic and stylised images. Currently the most preffered once are Flux 1.1 Pro and Flux 1.1 Ultra (better on follow detailed instructions, faster but more expensive).
Juggernaut
The juggernaut is an alternative to flux by RunDiffusion. Where:
- juggernaut base is cheap
- juggernaut lightning is fast
- juggernaut pro is the most advanced
Logos & Graphics
Ideogram
The ideogram is a text to image model designed to create graphic, typography and branding.
Product Photography
Bria
The bria is good at creating product shots.
Image Editing
Kontext
The kontext is using flux to edit/modify images
Upscale Image
Clarity Upscaler
The clarity-upscaler to increase resolution of an image. It is an alternative to magnific.
Train LoRA
LoRA (Low-Rank Adaptation) is a parameter-efficient fine-tuning (PEFT) technique that allows developers to customize large pre-trained models for specific tasks without retraining the entire model.
Flux LoRA Portrait Trainer
The flux lora portrait trainer is used to train the model by uploading images of a person, animal or object. After this you may use flux lora to generate new images (may be you would like to increase 'Num Inference Steps' from 28 to 40, if you don't like the result you may increase the 'Scale' form 1 to 1.2).
