Diffussion models - Revision history

Root at 08:25, 3 November 2025

2025-11-03T08:25:05Z

← Older revision		Revision as of 09:25, 3 November 2025
Line 60:		Line 60:

	= Using Difussion models locally =		= Using Difussion models locally =
	[[Install ~~Comfy UI~~\|Install ~~Comfy UI~~]]		[[Install ComfyUI\|Install ComfyUI]]

	[[AI: Text to image in local\|Text to image tutorial]]		[[AI: Text to image in local\|Text to image tutorial]]

Root at 08:24, 3 November 2025

2025-11-03T08:24:42Z

← Older revision		Revision as of 09:24, 3 November 2025
Line 60:		Line 60:

	= Using Difussion models locally =		= Using Difussion models locally =
	[Install Comfy UI\|Install Comfy UI]]		[[Install Comfy UI\|Install Comfy UI]]

	[[AI: Text to image in local\|Text to image tutorial]]		[[AI: Text to image in local\|Text to image tutorial]]

Root at 08:24, 3 November 2025

2025-11-03T08:24:22Z

← Older revision		Revision as of 09:24, 3 November 2025
Line 60:		Line 60:

	= Using Difussion models locally =		= Using Difussion models locally =
			[Install Comfy UI\|Install Comfy UI]]

	[[AI: Text to image in local\|Text to image tutorial]]		[[AI: Text to image in local\|Text to image tutorial]]

	[[AI: Image to video in local\|Image to video tutorial]]		[[AI: Image to video in local\|Image to video tutorial]]

	[[~~index.php?title=~~Category:KC AI technical aspects]]		[[Category:KC AI technical aspects]]

Root: /* VAE (Variational AutoEncoder) */

2025-11-03T08:23:15Z

VAE (Variational AutoEncoder)

← Older revision		Revision as of 09:23, 3 November 2025
Line 32:		Line 32:

	== VAE (Variational AutoEncoder) ==		== VAE (Variational AutoEncoder) ==
	~~Converts~~ images to and from compressed latent space, allowing diffusion models to work efficiently without directly processing full-resolution pixels.		A VAE is a tool that learns to shrink images into simple codes and then rebuild them, keeping their main details. It converts images to and from compressed latent space, allowing diffusion models to work efficiently without directly processing full-resolution pixels.

	A latent space is a compressed mathematical representation of an image where complex visual details are stored as simple numerical patterns~~, it captures~~ the essence of the image in fewer dimensions		A latent space is a compressed mathematical representation of an image, where complex visual details are encoded and stored as simple numerical patterns that capture the essence of the image in fewer dimensions.

	== Model sampling ==		== Model sampling ==
Line 64:		Line 64:
	[[AI: Image to video in local\|Image to video tutorial]]		[[AI: Image to video in local\|Image to video tutorial]]

	[[Category:KC AI technical aspects]]		[[index.php?title=Category:KC AI technical aspects]]

Root at 08:19, 3 November 2025

2025-11-03T08:19:39Z

← Older revision		Revision as of 09:19, 3 November 2025
Line 64:		Line 64:
	[[AI: Image to video in local\|Image to video tutorial]]		[[AI: Image to video in local\|Image to video tutorial]]

	[[~~index.php?title=~~Category:KC AI technical aspects]]		[[Category:KC AI technical aspects]]

Root: /* Quantification (Q) */

2025-11-03T08:18:23Z

Quantification (Q)

← Older revision		Revision as of 09:18, 3 November 2025
Line 21:		Line 21:
	They optimize internal calculations, allowing f'''ewer steps''' to reach '''similar quality'''.		They optimize internal calculations, allowing f'''ewer steps''' to reach '''similar quality'''.

	== ~~Quantification~~ (Q) ==		== Quantization (Q) ==
	If a model is quantified it is a '''compressed''' version of the original model where the numbers are stored with fewer bits (less precision). It saves memory and runs faster, at the cost of some quality.		If a model is quantified it is a '''compressed''' version of the original model where the numbers are stored with fewer bits (less precision). It saves memory and runs faster, at the cost of some quality. The number indicates how many bits per weight he model uses.

	The '''higher''' '''number''' of the ~~quantification~~ is, '''~~lower~~ quality''' the result will have and '''~~lower~~ memory''' requirements will have.		The '''higher''' '''number''' of the quantization is, '''higher quality''' the result will have and '''higher memory''' requirements will the model have.

	== Floating point (fp) ==		== Floating point (fp) ==
	It is a way of storing real numbers (like 3.14159) with '''high precision'''. Used for model weights and calculations.		It is a way of storing real numbers (like 3.14159) with '''high precision'''. Used for model weights and calculations.

	The '''higher number''' of fp is, '''higher quality''' will have the result and '''higher memory''' requirements will have.		The '''higher number''' of fp is, '''higher quality''' will have the result and '''higher memory''' requirements will the model have.

	== VAE (Variational AutoEncoder) ==		== VAE (Variational AutoEncoder) ==

Root at 07:51, 3 November 2025

2025-11-03T07:51:00Z

← Older revision		Revision as of 08:51, 3 November 2025
Line 48:		Line 48:
	== GGUF ==		== GGUF ==
	GGUF is a file format for large language models (LLMs) that optimizes them for consumer hardware by allowing them to be compressed and run efficiently on standard CPUs. They are converted and quantized versions of large diffusion or language models.		GGUF is a file format for large language models (LLMs) that optimizes them for consumer hardware by allowing them to be compressed and run efficiently on standard CPUs. They are converted and quantized versions of large diffusion or language models.

			== Checkpoint ==
			A checkpoint is a container. Inside it are the unet (gguf) model which is the one doing the actual diffusion, the clip encoder that translates your prompts and the vae that translates images to/from latent space.

	== Clip embeddings (comfyUI "loader" node) ==		== Clip embeddings (comfyUI "loader" node) ==
Line 61:		Line 64:
	[[AI: Image to video in local\|Image to video tutorial]]		[[AI: Image to video in local\|Image to video tutorial]]

	[[Category:KC AI technical aspects]]		[[index.php?title=Category:KC AI technical aspects]]

Root: Root moved page Difussion models to Diffussion models without leaving a redirect

2025-11-03T06:24:18Z

Root moved page Difussion models to Diffussion models without leaving a redirect

← Older revision	Revision as of 07:24, 3 November 2025
(No difference)

Root at 08:13, 31 October 2025

2025-10-31T08:13:19Z

← Older revision		Revision as of 09:13, 31 October 2025
Line 31:		Line 31:
	The '''higher number''' of fp is, '''higher quality''' will have the result and '''higher memory''' requirements will have.		The '''higher number''' of fp is, '''higher quality''' will have the result and '''higher memory''' requirements will have.

	== VAE ==		== VAE (Variational AutoEncoder) ==
	Converts images to and from compressed latent space, allowing diffusion models to work efficiently without directly processing full-resolution pixels.		Converts images to and from compressed latent space, allowing diffusion models to work efficiently without directly processing full-resolution pixels.

Root at 08:09, 31 October 2025

2025-10-31T08:09:56Z

← Older revision		Revision as of 09:09, 31 October 2025
Line 1:		Line 1:
	= What a difussion model is =		= What a difussion model is =
	It is a generative approach that learns to create data (like images or audio) by reversing a gradual noise-adding process. It starts with random noise and iteratively denoises it, guided by learned probabilities, to produce realistic samples that match the training distribution.		It is a generative approach that learns to create data (like images or audio) by reversing a gradual noise-adding process. It starts with random noise and iteratively denoises it, guided by learned probabilities, to produce realistic samples that match the training distribution.


	= Steps in the Diffusion Model process =		= Steps in the Diffusion Model process =
Line 8:		Line 9:
	# '''Guidance:''' Text or conditioning (via CLIP embeddings) directs the generation toward the desired prompt.		# '''Guidance:''' Text or conditioning (via CLIP embeddings) directs the generation toward the desired prompt.
	# '''Decoding:''' The final latent result is converted back to an image using the '''VAE'''.		# '''Decoding:''' The final latent result is converted back to an image using the '''VAE'''.
	= ~~basic~~ concepts =

			= Basic concepts =

	== Safetensor (LoRA) ==		== Safetensor (LoRA) ==