AI: Image to video in local - Revision history

Root at 21:15, 2 November 2025

2025-11-02T21:15:55Z

← Older revision		Revision as of 22:15, 2 November 2025
Line 1:		Line 1:
	TODO: create from text to video		TODO: create from text to video
	::		::
	[[Category:AI ~~Resources~~]]		[[Category:AI: Generate media locally]]

Root at 21:08, 2 November 2025

2025-11-02T21:08:58Z

← Older revision		Revision as of 22:08, 2 November 2025
Line 1:		Line 1:
	TODO: create from text to video		TODO: create from text to video
	::		::
			[[Category:AI Resources]]
	[[Category:IT Resources]]

Root: Replaced content with "TODO: create from text to video :: Category:IT Resources"

2025-10-31T08:03:03Z

Replaced content with "TODO: create from text to video :: Category:IT Resources"

@@ Line 1: / Line 1: @@
-= Basic concepts =
+TODO: create from text to video
 ::
 [[Category:IT Resources]]

Root at 23:21, 30 October 2025

2025-10-30T23:21:41Z

← Older revision		Revision as of 00:21, 31 October 2025
Line 1:		Line 1:
	= Basic concepts =		= Basic concepts =

	== ~~checkpoint~~ ==		== workflow ==
	~~A checkpoint is the original format for saving all model weights and configuration.~~

	=== noise level checkpoints ===		== noise level checkpoints ==
	For WAN model, there are high and low are different checkpoints of the WAN model that are designed to handle different noise levels or phases in the diffusion process (start from noise and gradually refine into frames).		For WAN model, there are high and low are different [[AI definitions#Checkpoint\|checkpoints]] of the WAN model that are designed to handle different noise levels or phases in the diffusion process (start from noise and gradually refine into frames).

	* A “high noise” model checkpoint is used for the '''earlier, high-noise phase''' of the generation process: when the latent is still very noisy, you need the model to do large corrections, heavy denoising, broad scene layout, motion establishment.		* A “high noise” model checkpoint is used for the '''earlier, high-noise phase''' of the generation process: when the latent is still very noisy, you need the model to do large corrections, heavy denoising, broad scene layout, motion establishment.
Line 15:		Line 14:

	== Safetensor ==		== Safetensor ==
	A safetensor is a small adapter that modify the main model’s weights at runtime (like LoRAs). It is a secure and fast file format for storing tensors, particularly large tensors like neural network weights.		A safetensor is a small adapter that modify the main model’s weights at runtime (like [[AI definitions#LoRA (Low-Rank Adaptation)\|LoRAs]]). It is a secure and fast file format for storing tensors, particularly large tensors like neural network weights.

	== Quantification (Q) ==		== Quantification (Q) ==
Line 26:		Line 25:

	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 ==

			== Clip loader ==

			== Model sampling ==

			== Sampler ==
	::		::

	[[Category:IT Resources]]		[[Category:IT Resources]]

Root at 22:58, 30 October 2025

2025-10-30T22:58:14Z

← Older revision		Revision as of 23:58, 30 October 2025
Line 1:		Line 1:
	= Basic concepts =		= Basic concepts =

			== checkpoint ==
			A checkpoint is the original format for saving all model weights and configuration.

			=== noise level checkpoints ===
			For WAN model, there are high and low are different checkpoints of the WAN model that are designed to handle different noise levels or phases in the diffusion process (start from noise and gradually refine into frames).

			* A “high noise” model checkpoint is used for the '''earlier, high-noise phase''' of the generation process: when the latent is still very noisy, you need the model to do large corrections, heavy denoising, broad scene layout, motion establishment.

			* A “low noise” model checkpoint is used for the '''later phase''', when the latent is already mostly refined and only fine-details, motion smoothing and subtle transitions remain.

	== GGUF ==		== GGUF ==
	It is a compressed ~~version of the model, with less parameters~~ and ~~quality loss~~.		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 the file~~ of ~~the model, it includes both the high and the low~~ models.

	== ~~safetensor~~ ==		== Safetensor ==
	It is ~~an extension where checkpoints work without compression~~.		A safetensor is a small adapter that modify the main model’s weights at runtime (like LoRAs). It is a secure and fast file format for storing tensors, particularly large tensors like neural network weights.

	== Quantification (Q) ==		== Quantification (Q) ==

Root: Root moved page AI: Image to image in local to AI: Image to video in local without leaving a redirect

2025-10-30T22:31:34Z

Root moved page AI: Image to image in local to AI: Image to video in local without leaving a redirect

← Older revision	Revision as of 23:31, 30 October 2025
(No difference)

Root at 13:00, 30 October 2025

2025-10-30T13:00:19Z

@@ Line 1: / Line 1: @@
-= Installing ComfyUI =
+= Basic concepts =
-= Understanding model compression and GPU VRAM =
+== GGUF ==
-== What is “compression” in the context of AI models? ==
+== checkpoint ==
-When we talk about compression for AI models (especially large video / image / generative models) we mean techniques that reduce the memory / compute cost required to run them, so they can fit on GPUs with less VRAM (video RAM) or run faster.
+A checkpoint is the file of the model, it includes both the high and the low models.
-Examples of compression techniques include:
+== safetensor ==
-* Quantisation (e.g., using 8-bit or 4-bit instead of 16- or 32-bit precision)
+It is an extension where checkpoints work without compression.
-== Why is VRAM important and how does it relate to model choice? ==
+== Quantification (Q) ==
-Your GPU’s VRAM (for example the VRAM on an NVIDIA RTX card) determines how large a model you can load and how much data (images, video frames) you can process at once.
+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.
-So when you pick a model, you must match:
+The '''higher''' '''number''' of the quantification is, '''lower quality''' the result will have and '''lower memory''' requirements will have.
-If you have less VRAM, you either need a smaller model (or more compressed version) or you have to reduce resolution/frame-count.
+== Floating point (fp) ==
-== Introducing the WAN 2.x models (text-to-video / image-to-video) ==
+The '''higher number''' of fp is, '''higher quality''' will have the result and '''higher memory''' requirements will have.
 ::
 [[Category:IT Resources]]

Root at 12:20, 30 October 2025

2025-10-30T12:20:25Z

@@ Line 1: / Line 1: @@
 = Installing ComfyUI =
-= Basic concepts =
+= Understanding model compression and GPU VRAM =
-Las versiones Q son las gguf, es el mismo modelo pero comprimido para hacer que pese menos y funcione en graficas con Vram mas pequeñas, pierdes algo de calidad, pero puedes usarlo en graficas mas malas ya que ocupan mucha menos memoria, las versiones Fp8 y fp16 son el modelo base tal cual, sin comprimir, solo que el fp8 tiene una compresion pero que apenas afecta a la calidad, son las mejores versiones del modelo sin perdida de calidad ni artefactos (apenas hay diferencia entre fp8 y fp16 y la diferencia es del doble)
+Examples of compression techniques include:
-Sobre el high y el low, es como esta entrenado Wan, dividieron el modelo en 2 pequeños modelos para hacerlo mas versatil y que funcione en mas equipos, en vez de tener 1 solo archivo de 30-50 gigas, como pasaba en wan 2.1, dividieron el modelo en 2, la teoria es que el modelo high se encarga del movimiento, el concepto base del video y la escena, y el low, refina los detalles finales, ambos son importantes.
+== Why is VRAM important and how does it relate to model choice? ==
-- Qué diferencia hay entre un gguf, checkpoint y safetensor?
+So when you pick a model, you must match:
-Un gguf, es la version comprimida del modelo,con menos parametros y perdida de calidad,lo que te dije arriba, necesitas un nodo que lea los gguf.
+If you have less VRAM, you either need a smaller model (or more compressed version) or you have to reduce resolution/frame-count.
-Un checkpoint es como comunmente llamamos al archivo del modelo solo que en ingles, el checkpoint de wan pues serian los dos archivos de high y low de decenas de gigas
+== Introducing the WAN 2.x models (text-to-video / image-to-video) ==
 [[Category:IT Resources]]

Root: Created page with "= Installing ComfyUI = = Basic concepts = Las versiones Q son las gguf, es el mismo modelo pero comprimido para hacer que pese menos y funcione en graficas con Vram mas pequeñas, pierdes algo de calidad, pero puedes usarlo en graficas mas malas ya que ocupan mucha menos memoria, las versiones Fp8 y fp16 son el modelo base tal cual, sin comprimir, solo que el fp8 tiene una compresion pero que apenas afecta a la calidad, son las mejores versiones del modelo sin perdida..."

2025-10-30T12:13:03Z

Created page with "= Installing ComfyUI = = Basic concepts = Las versiones Q son las gguf, es el mismo modelo pero comprimido para hacer que pese menos y funcione en graficas con Vram mas pequeñas, pierdes algo de calidad, pero puedes usarlo en graficas mas malas ya que ocupan mucha menos memoria, las versiones Fp8 y fp16 son el modelo base tal cual, sin comprimir, solo que el fp8 tiene una compresion pero que apenas afecta a la calidad, son las mejores versiones del modelo sin perdida..."

New page

= Installing ComfyUI =

= Basic concepts =

Las versiones Q son las gguf, es el mismo modelo pero comprimido para hacer que pese menos y funcione en graficas con Vram mas pequeñas, pierdes algo de calidad, pero puedes usarlo en graficas mas malas ya que ocupan mucha menos memoria, las versiones Fp8 y fp16 son el modelo base tal cual, sin comprimir, solo que el fp8 tiene una compresion pero que apenas afecta a la calidad, son las mejores versiones del modelo sin perdida de calidad ni artefactos (apenas hay diferencia entre fp8 y fp16 y la diferencia es del doble)

Sobre el high y el low, es como esta entrenado Wan, dividieron el modelo en 2 pequeños modelos para hacerlo mas versatil y que funcione en mas equipos, en vez de tener 1 solo archivo de 30-50 gigas, como pasaba en wan 2.1, dividieron el modelo en 2, la teoria es que el modelo high se encarga del movimiento, el concepto base del video y la escena, y el low, refina los detalles finales, ambos son importantes.

- Qué diferencia hay entre un gguf, checkpoint y safetensor?

Un gguf, es la version comprimida del modelo,con menos parametros y perdida de calidad,lo que te dije arriba, necesitas un nodo que lea los gguf.

Un checkpoint es como comunmente llamamos al archivo del modelo solo que en ingles, el checkpoint de wan pues serian los dos archivos de high y low de decenas de gigas

Y safetensors es la extension en la que funcionan los checkpoints sin comprimir.
[[Category:IT Resources]]