LTX-2.3: texto a vídeo
Genera un vídeo a partir de un prompt de texto, usando opcionalmente una imagen como referencia. Recibe un vídeo de alta calidad con movimiento, audio y detalle mejorados, optimizado para formatos vertical u horizontal.
Dentro del flujo de trabajo
Una vista previa de solo lectura de la plantilla de origen. Ábrela en Cloud para explorar los nodos y hacer cambios.
Loading the graph
Open full-size workflow previewLlama a este modelo desde tu código
La solicitud a continuación utiliza el mismo grafo preparado y la misma configuración que el Playground. Cada envío puede iniciar una ejecución de pago; no reintentes automáticamente un envío cuyo resultado se desconozca.
What you need
- Runs on
- Comfy Cloud
- Clave
- COMFY_API_KEY
Requiere un plan de pago de Cloud y créditos.
# Python 3.10+: pip install comfy-sdk==0.4.0
import os
from comfy_sdk import Comfy
api_key = os.environ.get("COMFY_API_KEY", "").strip()
if not api_key: raise RuntimeError("Set COMFY_API_KEY")
os.environ.setdefault("COMFY_BASE_URL", "https://cloud.comfy.org")
client = Comfy(api_key=api_key)
workflow = client.workflows.from_json({
"75": {
"class_type": "SaveVideo",
"inputs": {
"filename_prefix": "video/LTX_2.3_t2v",
"format": "auto",
"format.codec": "auto",
"video": [
"267:242",
0
]
},
"_meta": {
"title": "Save Video"
}
},
"269": {
"class_type": "LoadImage",
"inputs": {
"image": "example.png"
},
"_meta": {
"title": "Load Image"
}
},
"267:201": {
"class_type": "PrimitiveBoolean",
"inputs": {
"value": True
},
"_meta": {
"title": "Switch to Text to Video?"
}
},
"267:209": {
"class_type": "KSamplerSelect",
"inputs": {
"sampler_name": "euler_ancestral_cfg_pp"
},
"_meta": {
"title": "KSamplerSelect"
}
},
"267:211": {
"class_type": "ManualSigmas",
"inputs": {
"sigmas": "0.85, 0.7250, 0.4219, 0.0"
},
"_meta": {
"title": "ManualSigmas"
}
},
"267:212": {
"class_type": "LTXVCropGuides",
"inputs": {
"latent": [
"267:217",
0
],
"negative": [
"267:239",
1
],
"positive": [
"267:239",
0
]
},
"_meta": {
"title": "LTXVCropGuides"
}
},
"267:213": {
"class_type": "CFGGuider",
"inputs": {
"cfg": 1,
"model": [
"267:232",
0
],
"negative": [
"267:212",
1
],
"positive": [
"267:212",
0
]
},
"_meta": {
"title": "CFG Guider"
}
},
"267:214": {
"class_type": "LTXVEmptyLatentAudio",
"inputs": {
"audio_vae": [
"267:221",
0
],
"batch_size": 1,
"frame_rate": [
"267:261",
1
],
"frames_number": [
"267:225",
0
]
},
"_meta": {
"title": "Empty LTXV Latent Audio"
}
},
"267:215": {
"class_type": "SamplerCustomAdvanced",
"inputs": {
"guider": [
"267:231",
0
],
"latent_image": [
"267:222",
0
],
"noise": [
"267:237",
0
],
"sampler": [
"267:209",
0
],
"sigmas": [
"267:252",
0
]
},
"_meta": {
"title": "SamplerCustomAdvanced"
}
},
"267:216": {
"class_type": "RandomNoise",
"inputs": {
"noise_seed": 42
},
"_meta": {
"title": "RandomNoise"
}
},
"267:217": {
"class_type": "LTXVSeparateAVLatent",
"inputs": {
"av_latent": [
"267:215",
0
]
},
"_meta": {
"title": "Separate AV Latent"
}
},
"267:218": {
"class_type": "LTXVSeparateAVLatent",
"inputs": {
"av_latent": [
"267:219",
0
]
},
"_meta": {
"title": "Separate AV Latent"
}
},
"267:219": {
"class_type": "SamplerCustomAdvanced",
"inputs": {
"guider": [
"267:213",
0
],
"latent_image": [
"267:229",
0
],
"noise": [
"267:216",
0
],
"sampler": [
"267:246",
0
],
"sigmas": [
"267:211",
0
]
},
"_meta": {
"title": "SamplerCustomAdvanced"
}
},
"267:220": {
"class_type": "LTXVAudioVAEDecode",
"inputs": {
"audio_vae": [
"267:221",
0
],
"samples": [
"267:218",
1
]
},
"_meta": {
"title": "LTXV Audio VAE Decode"
}
},
"267:221": {
"class_type": "LTXVAudioVAELoader",
"inputs": {
"ckpt_name": "ltx-2.3-22b-dev-fp8.safetensors"
},
"_meta": {
"title": "Load LTXV Audio VAE"
}
},
"267:222": {
"class_type": "LTXVConcatAVLatent",
"inputs": {
"audio_latent": [
"267:214",
0
],
"video_latent": [
"267:249",
0
]
},
"_meta": {
"title": "Concat AV Latent"
}
},
"267:225": {
"class_type": "PrimitiveInt",
"inputs": {
"value": 121
},
"_meta": {
"title": "Length"
}
},
"267:228": {
"class_type": "EmptyLTXVLatentVideo",
"inputs": {
"batch_size": 1,
"height": [
"267:259",
1
],
"length": [
"267:225",
0
],
"width": [
"267:256",
1
]
},
"_meta": {
"title": "Empty LTXV Latent Video"
}
},
"267:229": {
"class_type": "LTXVConcatAVLatent",
"inputs": {
"audio_latent": [
"267:217",
1
],
"video_latent": [
"267:230",
0
]
},
"_meta": {
"title": "Concat AV Latent"
}
},
"267:230": {
"class_type": "LTXVImgToVideoInplace",
"inputs": {
"bypass": [
"267:201",
0
],
"image": [
"267:248",
0
],
"latent": [
"267:253",
0
],
"strength": 1,
"vae": [
"267:236",
2
]
},
"_meta": {
"title": "LTXVImgToVideoInplace"
}
},
"267:231": {
"class_type": "CFGGuider",
"inputs": {
"cfg": 1,
"model": [
"267:232",
0
],
"negative": [
"267:239",
1
],
"positive": [
"267:239",
0
]
},
"_meta": {
"title": "CFG Guider"
}
},
"267:232": {
"class_type": "LoraLoaderModelOnly",
"inputs": {
"lora_name": "ltx-2.3-22b-distilled-lora-384.safetensors",
"model": [
"267:236",
0
],
"strength_model": 0.5
},
"_meta": {
"title": "Load LoRA"
}
},
"267:233": {
"class_type": "LatentUpscaleModelLoader",
"inputs": {
"model_name": "ltx-2.3-spatial-upscaler-x2-1.1.safetensors"
},
"_meta": {
"title": "Load Latent Upscale Model"
}
},
"267:235": {
"class_type": "ResizeImagesByLongerEdge",
"inputs": {
"images": [
"267:238",
0
],
"longer_edge": 1536
},
"_meta": {
"title": "Resize Images by Longer Edge (DEPRECATED)"
}
},
"267:236": {
"class_type": "CheckpointLoaderSimple",
"inputs": {
"ckpt_name": "ltx-2.3-22b-dev-fp8.safetensors"
},
"_meta": {
"title": "Load Checkpoint"
}
},
"267:237": {
"class_type": "RandomNoise",
"inputs": {
"noise_seed": 731987174230470
},
"_meta": {
"title": "RandomNoise"
}
},
"267:238": {
"class_type": "ResizeImageMaskNode",
"inputs": {
"input": [
"269",
0
],
"resize_type": "scale dimensions",
"resize_type.crop": "center",
"resize_type.height": [
"267:258",
0
],
"resize_type.width": [
"267:257",
0
],
"scale_method": "lanczos"
},
"_meta": {
"title": "Resize Image/Mask"
}
},
"267:239": {
"class_type": "LTXVConditioning",
"inputs": {
"frame_rate": [
"267:261",
0
],
"negative": [
"267:247",
0
],
"positive": [
"267:240",
0
]
},
"_meta": {
"title": "LTXVConditioning"
}
},
"267:240": {
"class_type": "CLIPTextEncode",
"inputs": {
"clip": [
"267:243",
0
],
"text": [
"267:274",
0
]
},
"_meta": {
"title": "CLIP Text Encode (Prompt)"
}
},
"267:242": {
"class_type": "CreateVideo",
"inputs": {
"audio": [
"267:220",
0
],
"bit_depth": "auto",
"codec": "none",
"color_space": "sRGB",
"fps": [
"267:261",
0
],
"images": [
"267:251",
0
]
},
"_meta": {
"title": "Create Video"
}
},
"267:243": {
"class_type": "LTXAVTextEncoderLoader",
"inputs": {
"ckpt_name": "ltx-2.3-22b-dev-fp8.safetensors",
"device": "default",
"text_encoder": "gemma_3_12B_it_fp4_mixed.safetensors"
},
"_meta": {
"title": "Load LTXV Audio Text Encoder"
}
},
"267:246": {
"class_type": "KSamplerSelect",
"inputs": {
"sampler_name": "euler_cfg_pp"
},
"_meta": {
"title": "KSamplerSelect"
}
},
"267:247": {
"class_type": "CLIPTextEncode",
"inputs": {
"clip": [
"267:272",
1
],
"text": "pc game, console game, video game, cartoon, childish, ugly"
},
"_meta": {
"title": "CLIP Text Encode (Prompt)"
}
},
"267:248": {
"class_type": "LTXVPreprocess",
"inputs": {
"image": [
"267:235",
0
],
"img_compression": 18
},
"_meta": {
"title": "LTXV Preprocess"
}
},
"267:249": {
"class_type": "LTXVImgToVideoInplace",
"inputs": {
"bypass": [
"267:201",
0
],
"image": [
"267:248",
0
],
"latent": [
"267:228",
0
],
"strength": 0.7,
"vae": [
"267:236",
2
]
},
"_meta": {
"title": "LTXVImgToVideoInplace"
}
},
"267:251": {
"class_type": "VAEDecodeTiled",
"inputs": {
"overlap": 64,
"samples": [
"267:218",
0
],
"temporal_overlap": 4,
"temporal_size": 4096,
"tile_size": 768,
"vae": [
"267:236",
2
]
},
"_meta": {
"title": "VAE Decode (Tiled)"
}
},
"267:252": {
"class_type": "ManualSigmas",
"inputs": {
"sigmas": "1.0, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875, 0.0"
},
"_meta": {
"title": "ManualSigmas"
}
},
"267:253": {
"class_type": "LTXVLatentUpsampler",
"inputs": {
"samples": [
"267:217",
0
],
"upscale_model": [
"267:233",
0
],
"vae": [
"267:236",
2
]
},
"_meta": {
"title": "LTXVLatentUpsampler"
}
},
"267:256": {
"class_type": "ComfyMathExpression",
"inputs": {
"expression": "a/2",
"values.a": [
"267:257",
0
]
},
"_meta": {
"title": "Math Expression"
}
},
"267:257": {
"class_type": "PrimitiveInt",
"inputs": {
"value": 1280
},
"_meta": {
"title": "Width"
}
},
"267:258": {
"class_type": "PrimitiveInt",
"inputs": {
"value": 720
},
"_meta": {
"title": "Height"
}
},
"267:259": {
"class_type": "ComfyMathExpression",
"inputs": {
"expression": "a/2",
"values.a": [
"267:258",
0
]
},
"_meta": {
"title": "Math Expression"
}
},
"267:260": {
"class_type": "PrimitiveInt",
"inputs": {
"value": 25
},
"_meta": {
"title": "Frame Rate"
}
},
"267:261": {
"class_type": "ComfyMathExpression",
"inputs": {
"expression": "a",
"values.a": [
"267:260",
0
]
},
"_meta": {
"title": "Math Expression"
}
},
"267:266": {
"class_type": "PrimitiveStringMultiline",
"inputs": {
"value": "At stormy dusk, a solitary medieval warrior clad in hammered iron plate and sodden chainmail stands atop an expansive, muddy ridge. He wears an open-faced, rounded steel helmet revealing a weathered, scarred face streaked with grime. Sheets of heavy rain cut diagonally through the air, catching the fiery amber glow of a setting sun trapped beneath bruised violet storm clouds. Over five seconds, gale-force winds violently whip his frayed crimson cloak while tempestuous clouds churn rapidly overhead. The warrior slowly shifts his weight, gripping a notched steel longsword. The camera executes a low, continuous arc around his waist before smoothly pushing into a dramatic close-up of his steely, rain-soaked gaze. Ambient audio carries the violent drumming of downpour against steel plate, howling wind, and the distinct, wet creak of shifting leather armor."
},
"_meta": {
"title": "Prompt"
}
},
"267:272": {
"class_type": "LoraLoader",
"inputs": {
"clip": [
"267:243",
0
],
"lora_name": "gemma-3-12b-it-abliterated_lora_rank64_bf16.safetensors",
"model": [
"267:236",
0
],
"strength_clip": 1,
"strength_model": 1
},
"_meta": {
"title": "Load LoRA (Model and CLIP)"
}
},
"267:274": {
"class_type": "TextGenerateLTX2Prompt",
"inputs": {
"clip": [
"267:272",
1
],
"max_length": 256,
"mtp": "auto",
"prompt": [
"267:266",
0
],
"sampling_mode": "on",
"sampling_mode.min_p": 0.05,
"sampling_mode.repetition_penalty": 1.05,
"sampling_mode.seed": 0,
"sampling_mode.temperature": 0.7,
"sampling_mode.top_k": 64,
"sampling_mode.top_p": 0.95,
"thinking": False,
"use_default_template": True
},
"_meta": {
"title": "Generate LTX2 Prompt"
}
},
"267:275": {
"class_type": "PreviewAny",
"inputs": {
"source": [
"267:274",
0
]
},
"_meta": {
"title": "Preview as Text"
}
}
})
# run() also passes api_key to the partner nodes in this workflow.
job = client.run(workflow, api_key=api_key)
for output in job.get_outputs("75"):
output.to_file(output.name)Probar un ejemplo
Acerca de este flujo de trabajo
El flujo de trabajo LTX-2.3: Texto a Vídeo en ComfyUI está diseñado para transformar un prompt de texto en un vídeo dinámico, con la opción de usar una imagen como referencia. Este flujo de trabajo aprovecha el modelo LTX-2.3, desarrollado por Lightricks, optimizado para generar vídeos de alta calidad con movimiento, audio y detalle mejorados. El flujo de trabajo es especialmente eficaz para crear vídeos tanto en formato vertical como horizontal, lo que lo hace versátil para distintas aplicaciones.
Técnicamente, el flujo de trabajo integra varios nodos clave, incluidos el nodo SaveVideo para generar el vídeo final y el nodo LoadImage para incorporar imágenes de referencia. El nodo MarkdownNote se usa para ofrecer documentación en línea y orientación dentro del flujo de trabajo. El núcleo del proceso consiste en usar el nodo b94257db-cdc1-45d3-8913-ca61e782d9c1, responsable de interpretar el prompt de texto y generar el contenido de vídeo correspondiente. Este flujo de trabajo es útil para usuarios que buscan producir rápidamente vídeos visualmente atractivos a partir de descripciones textuales, con la capacidad adicional de afinar la salida usando imágenes de referencia.
Preguntas y respuestas
Sí, es necesario subir una imagen para proporcionar una referencia visual y mejorar el detalle en el vídeo generado.
Sí, el flujo de trabajo está optimizado para formato vertical y horizontal, para que puedas elegir según tus necesidades.
El flujo de trabajo usa el modelo LTX-2.3 desarrollado por Lightricks, disponible en Hugging Face.
Puedes refinar el vídeo ajustando el prompt de texto para descripciones más precisas y subiendo distintas imágenes de referencia para influir en los detalles visuales.