Restaurer des Archives Vidéo - LTX 2.3 Dearchive LoRA
Téléchargez une vidéo ou un extrait d'archive et restaurez-le en 1 clic.
Dans le workflow
Un aperçu en lecture seule du modèle source. Ouvrez-le dans Cloud pour explorer les nœuds et apporter des modifications.
Loading the graph
Open full-size workflow previewFichiers de ce workflow
- Sortie
- Vidéo
- Auteur
- Rob
Appelez ce modèle depuis votre code
La requête ci-dessous utilise le même graphe préparé et les mêmes paramètres que le Playground. Chaque envoi peut lancer une exécution payante ; ne relancez pas automatiquement un envoi dont l’issue est inconnue.
What you need
- Runs on
- Comfy Cloud
- Clé
- COMFY_API_KEY
- Vos fichiers
- Importés par le code depuis leurs URL
Nécessite un forfait Cloud payant et des crédits.
# 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({
"5187": {
"class_type": "SaveVideo",
"inputs": {
"filename_prefix": "video",
"format": "auto",
"format.codec": "auto",
"video": [
"5189:5062",
0
]
},
"_meta": {
"title": "Save Video"
}
},
"5188": {
"class_type": "LoadVideo",
"inputs": {
"file": "UPLOADED_video_FILENAME"
},
"_meta": {
"title": "Load Video"
}
},
"5190": {
"class_type": "GetVideoComponents",
"inputs": {
"video": [
"5188",
0
]
},
"_meta": {
"title": "Get Video Components"
}
},
"5192": {
"class_type": "SaveVideo",
"inputs": {
"filename_prefix": "video_compare",
"format": "auto",
"format.codec": "auto",
"video": [
"5191:80",
0
]
},
"_meta": {
"title": "Save Comparison Video"
}
},
"5189:5057": {
"class_type": "CLIPTextEncode",
"inputs": {
"clip": [
"5189:5084",
0
],
"text": "pc game, console game, video game, cartoon, childish, ugly"
},
"_meta": {
"title": "CLIP Text Encode (Negative Prompt)"
}
},
"5189:5058": {
"class_type": "RandomNoise",
"inputs": {
"noise_seed": 234967385442993
},
"_meta": {
"title": "RandomNoise"
}
},
"5189:5059": {
"class_type": "LatentUpscaleModelLoader",
"inputs": {
"model_name": "ltx-2.3-spatial-upscaler-x2-1.1.safetensors"
},
"_meta": {
"title": "Load Latent Upscale Model"
}
},
"5189:5060": {
"class_type": "LTXVConditioning",
"inputs": {
"frame_rate": [
"5189:5088",
2
],
"negative": [
"5189:5057",
0
],
"positive": [
"5189:5091",
0
]
},
"_meta": {
"title": "LTXVConditioning"
}
},
"5189:5061": {
"class_type": "CFGGuider",
"inputs": {
"cfg": 1,
"model": [
"5189:5087",
0
],
"negative": [
"5189:5098",
1
],
"positive": [
"5189:5098",
0
]
},
"_meta": {
"title": "CFG Guider"
}
},
"5189:5062": {
"class_type": "CreateVideo",
"inputs": {
"audio": [
"5190",
1
],
"bit_depth": "auto",
"codec": "none",
"color_space": "sRGB",
"fps": [
"5189:5088",
2
],
"images": [
"5189:5090",
0
]
},
"_meta": {
"title": "Create Video"
}
},
"5189:5063": {
"class_type": "KSamplerSelect",
"inputs": {
"sampler_name": "euler"
},
"_meta": {
"title": "KSamplerSelect"
}
},
"5189:5064": {
"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"
}
},
"5189:5065": {
"class_type": "LTXVEmptyLatentAudio",
"inputs": {
"audio_vae": [
"5189:5086",
0
],
"batch_size": 1,
"frame_rate": [
"5189:5095",
0
],
"frames_number": [
"5189:5097",
1
]
},
"_meta": {
"title": "Empty LTXV Latent Audio"
}
},
"5189:5066": {
"class_type": "LTXVConcatAVLatent",
"inputs": {
"audio_latent": [
"5189:5065",
0
],
"video_latent": [
"5189:5094",
2
]
},
"_meta": {
"title": "Concat AV Latent"
}
},
"5189:5067": {
"class_type": "SamplerCustomAdvanced",
"inputs": {
"guider": [
"5189:5061",
0
],
"latent_image": [
"5189:5066",
0
],
"noise": [
"5189:5058",
0
],
"sampler": [
"5189:5063",
0
],
"sigmas": [
"5189:5064",
0
]
},
"_meta": {
"title": "SamplerCustomAdvanced"
}
},
"5189:5068": {
"class_type": "ResizeImageMaskNode",
"inputs": {
"input": [
"5189:5088",
0
],
"resize_type": "scale dimensions",
"resize_type.crop": "disabled",
"resize_type.height": 1088,
"resize_type.width": 1920,
"scale_method": "area"
},
"_meta": {
"title": "Resize Image/Mask (s2 size)"
}
},
"5189:5069": {
"class_type": "LTXVLatentUpsampler",
"inputs": {
"samples": [
"5189:5082",
2
],
"upscale_model": [
"5189:5059",
0
],
"vae": [
"5189:5085",
2
]
},
"_meta": {
"title": "LTXVLatentUpsampler"
}
},
"5189:5070": {
"class_type": "LTXVSeparateAVLatent",
"inputs": {
"av_latent": [
"5189:5067",
0
]
},
"_meta": {
"title": "Separate AV Latent"
}
},
"5189:5071": {
"class_type": "LTXAddVideoICLoRAGuide",
"inputs": {
"crop": "disabled",
"frame_idx": 0,
"image": [
"5189:5068",
0
],
"latent": [
"5189:5069",
0
],
"latent_downscale_factor": 1,
"negative": [
"5189:5099",
1
],
"positive": [
"5189:5099",
0
],
"strength": 1,
"tile_frames": 97,
"tile_overlap": 64,
"tile_size": 256,
"use_streaming": False,
"use_tiled_encode": True,
"vae": [
"5189:5085",
2
]
},
"_meta": {
"title": "🅛🅣🅧 Add Video IC-LoRA Guide"
}
},
"5189:5072": {
"class_type": "KSamplerSelect",
"inputs": {
"sampler_name": "euler"
},
"_meta": {
"title": "KSamplerSelect"
}
},
"5189:5073": {
"class_type": "SamplerCustomAdvanced",
"inputs": {
"guider": [
"5189:5076",
0
],
"latent_image": [
"5189:5077",
0
],
"noise": [
"5189:5078",
0
],
"sampler": [
"5189:5072",
0
],
"sigmas": [
"5189:5075",
0
]
},
"_meta": {
"title": "SamplerCustomAdvanced"
}
},
"5189:5074": {
"class_type": "LTXVCropGuides",
"inputs": {
"latent": [
"5189:5080",
0
],
"negative": [
"5189:5071",
1
],
"positive": [
"5189:5071",
0
]
},
"_meta": {
"title": "LTXVCropGuides"
}
},
"5189:5075": {
"class_type": "ManualSigmas",
"inputs": {
"sigmas": "0.909375, 0.725, 0.421875, 0.0"
},
"_meta": {
"title": "ManualSigmas"
}
},
"5189:5076": {
"class_type": "CFGGuider",
"inputs": {
"cfg": 1,
"model": [
"5189:5087",
0
],
"negative": [
"5189:5071",
1
],
"positive": [
"5189:5071",
0
]
},
"_meta": {
"title": "CFG Guider"
}
},
"5189:5077": {
"class_type": "LTXVConcatAVLatent",
"inputs": {
"audio_latent": [
"5189:5099",
2
],
"video_latent": [
"5189:5071",
2
]
},
"_meta": {
"title": "Concat AV Latent"
}
},
"5189:5078": {
"class_type": "RandomNoise",
"inputs": {
"noise_seed": 29
},
"_meta": {
"title": "RandomNoise"
}
},
"5189:5079": {
"class_type": "LTXVAudioVAEDecode",
"inputs": {
"audio_vae": [
"5189:5086",
0
],
"samples": [
"5189:5080",
1
]
},
"_meta": {
"title": "LTXV Audio VAE Decode"
}
},
"5189:5080": {
"class_type": "LTXVSeparateAVLatent",
"inputs": {
"av_latent": [
"5189:5073",
0
]
},
"_meta": {
"title": "Separate AV Latent"
}
},
"5189:5082": {
"class_type": "LTXVCropGuides",
"inputs": {
"latent": [
"5189:5070",
0
],
"negative": [
"5189:5094",
1
],
"positive": [
"5189:5094",
0
]
},
"_meta": {
"title": "LTXVCropGuides"
}
},
"5189:5083": {
"class_type": "LTXVAudioVAEEncode",
"inputs": {
"audio": [
"5189:5088",
1
],
"audio_vae": [
"5189:5086",
0
]
},
"_meta": {
"title": "LTXV Audio VAE Encode"
}
},
"5189:5084": {
"class_type": "LTXAVTextEncoderLoader",
"inputs": {
"ckpt_name": "ltx-2.3-22b-dev.safetensors",
"device": "default",
"text_encoder": "gemma_3_12B_it.safetensors"
},
"_meta": {
"title": "Load LTXV Audio Text Encoder"
}
},
"5189:5085": {
"class_type": "CheckpointLoaderSimple",
"inputs": {
"ckpt_name": "ltx-2.3-22b-dev.safetensors"
},
"_meta": {
"title": "Load Checkpoint"
}
},
"5189:5086": {
"class_type": "LTXVAudioVAELoader",
"inputs": {
"ckpt_name": "ltx-2.3-22b-dev.safetensors"
},
"_meta": {
"title": "Load LTXV Audio VAE"
}
},
"5189:5087": {
"class_type": "LTXICLoRALoaderModelOnly",
"inputs": {
"lora_name": "ltx-2.3-dearchive-lora_weights_step_05000.safetensors",
"model": [
"5189:5085",
0
],
"strength_model": 1
},
"_meta": {
"title": "🅛🅣🅧 IC-LoRA Loader Model Only"
}
},
"5189:5088": {
"class_type": "GetVideoComponents",
"inputs": {
"video": [
"5188",
0
]
},
"_meta": {
"title": "Get Video Components"
}
},
"5189:5089": {
"class_type": "ResizeImageMaskNode",
"inputs": {
"input": [
"5189:5088",
0
],
"resize_type": "scale dimensions",
"resize_type.crop": "disabled",
"resize_type.height": 544,
"resize_type.width": 960,
"scale_method": "area"
},
"_meta": {
"title": "Resize Image/Mask (s1 size)"
}
},
"5189:5090": {
"class_type": "LTXVTiledVAEDecode",
"inputs": {
"horizontal_tiles": 2,
"last_frame_fix": False,
"latents": [
"5189:5074",
2
],
"overlap": 6,
"vae": [
"5189:5085",
2
],
"vertical_tiles": 2,
"working_device": "auto",
"working_dtype": "auto"
},
"_meta": {
"title": "🅛🅣🅧 LTXV Tiled VAE Decode"
}
},
"5189:5091": {
"class_type": "CLIPTextEncode",
"inputs": {
"clip": [
"5189:5084",
0
],
"text": "A modern, high-resolution colour video of a busy 1920s harbour street, shot with sharp detail and contemporary cinematography. Dock workers in faded blue cotton tunics and brown canvas trousers haul pale burlap sacks across grey granite cobbles, a cream and maroon streetcar rolls along polished steel rails, and an elderly man in a long black silk robe walks calmly through the crowd. Red brick warehouses with green canvas awnings line the street, wooden crates and coiled ropes sit beside black iron bollards, and tugboats with red funnels release white steam over the blue-grey water. Clean, stable frames with natural skin tones, soft daylight from an overcast sky, crisp edges, no grain, no scratches, no flicker."
},
"_meta": {
"title": "CLIP Text Encode (Positive Prompt)"
}
},
"5189:5093": {
"class_type": "EmptyLTXVLatentVideo",
"inputs": {
"batch_size": 1,
"height": 544,
"length": [
"5189:5097",
1
],
"width": 960
},
"_meta": {
"title": "Empty LTXV Latent Video"
}
},
"5189:5094": {
"class_type": "LTXAddVideoICLoRAGuide",
"inputs": {
"crop": "disabled",
"frame_idx": 0,
"image": [
"5189:5089",
0
],
"latent": [
"5189:5093",
0
],
"latent_downscale_factor": 1,
"negative": [
"5189:5060",
1
],
"positive": [
"5189:5060",
0
],
"strength": 1,
"tile_frames": 97,
"tile_overlap": 64,
"tile_size": 256,
"use_streaming": False,
"use_tiled_encode": True,
"vae": [
"5189:5085",
2
]
},
"_meta": {
"title": "🅛🅣🅧 Add Video IC-LoRA Guide"
}
},
"5189:5095": {
"class_type": "LTXFloatToInt",
"inputs": {
"a": [
"5189:5088",
2
]
},
"_meta": {
"title": "🅛🅣🅧 Float To Int"
}
},
"5189:5096": {
"class_type": "GetImageSizeAndCount",
"inputs": {
"image": [
"5189:5089",
0
]
},
"_meta": {
"title": "Get Image Size & Count"
}
},
"5189:5097": {
"class_type": "ComfyMathExpression",
"inputs": {
"expression": "(int((a - 1) / 8)) * 8 + 1",
"values.a": [
"5189:5096",
3
]
},
"_meta": {
"title": "Math Expression"
}
},
"5189:5098": {
"class_type": "LTXVSetAudioRefTokens",
"inputs": {
"audio_latent": [
"5189:5083",
0
],
"negative": [
"5189:5094",
1
],
"positive": [
"5189:5094",
0
]
},
"_meta": {
"title": "🅛🅣🅧 Set Audio Ref Tokens"
}
},
"5189:5099": {
"class_type": "LTXVSetAudioRefTokens",
"inputs": {
"audio_latent": [
"5189:5070",
1
],
"negative": [
"5189:5082",
1
],
"positive": [
"5189:5082",
0
]
},
"_meta": {
"title": "🅛🅣🅧 Set Audio Ref Tokens"
}
},
"5191:77": {
"class_type": "GetVideoComponents",
"inputs": {
"video": [
"5188",
0
]
},
"_meta": {
"title": "Get Video Components"
}
},
"5191:78": {
"class_type": "GetVideoComponents",
"inputs": {
"video": [
"5189:5062",
0
]
},
"_meta": {
"title": "Get Video Components"
}
},
"5191:79": {
"class_type": "ImageStitch",
"inputs": {
"direction": "down",
"image1": [
"5191:77",
0
],
"image2": [
"5191:78",
0
],
"match_image_size": False,
"spacing_color": "white",
"spacing_width": 0
},
"_meta": {
"title": "Stitch Images"
}
},
"5191:80": {
"class_type": "CreateVideo",
"inputs": {
"audio": [
"5191:77",
1
],
"bit_depth": "auto",
"codec": "none",
"color_space": "sRGB",
"fps": [
"5191:77",
2
],
"images": [
"5191:79",
0
]
},
"_meta": {
"title": "Create Video"
}
}
})
workflow.set_input("5188", "file", client.assets.from_url("https://comfy.org/catalog-media/1c731953-3989-4103-ab64-6e34cd226332/73ad917e-c935-4746-be03-2c478b979f32"))
# 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("5187"):
output.to_file(output.name)
for output in job.get_outputs("5192"):
output.to_file(output.name)Essayer un exemple
À propos de ce workflow
Ce workflow restaure des vidéos d'archives ou dégradées en un seul passage grâce à un backbone LTX 2.3 guidé par un Dearchive LoRA. Il commence par décoder votre clip avec LoadVideo et lire les données de timing/taille via GetVideoComponents, puis achemine les images à travers le cœur de restauration avant de réencoder en MP4 propre avec SaveVideo. Les réglages par défaut sont optimisés pour réduire les artefacts courants des archives comme le grain, le scintillement, les rayures et le bruit de compression, tout en conservant le cadrage et le timing d'origine.
Techniquement, deux nœuds personnalisés (IDs : 14beb77c-6207-496b-9559-1eeb98b2ee76 et 5784f1cc-cf8c-4289-9c5b-2e837048198f) implémentent le passage modèle LTX 2.3 et appliquent le Dearchive LoRA. GetVideoComponents transmet la fréquence et la résolution pour que SaveVideo encode au bon timing, assurant la cohérence temporelle. Si les nœuds personnalisés exposent un paramètre de poids/force LoRA, vous pouvez ajuster le niveau de restauration : des valeurs plus élevées accentuent le nettoyage, des valeurs plus basses conservent plus de texture originale.
Questions-réponses
La restauration est réalisée par deux nœuds personnalisés dans ce graphe (IDs : 14beb77c-6207-496b-9559-1eeb98b2ee76 et 5784f1cc-cf8c-4289-9c5b-2e837048198f). Ils exécutent le modèle LTX 2.3 avec un Dearchive LoRA pour nettoyer les images. LoadVideo/GetVideoComponents gèrent le décodage et le timing ; SaveVideo encode le résultat.
SaveVideo de ComfyUI exporte généralement sans audio sauf si le graphe achemine explicitement l'audio. Ce modèle utilise GetVideoComponents pour FPS/taille ; si l'audio n'est pas connecté à SaveVideo, l'export sera muet. Pour garder l'audio, remuxez la piste originale avec la vidéo nettoyée via ffmpeg, ou ajoutez un passage audio dans le graphe.
Si le nœud de restauration expose un paramètre de force/poids LoRA, des valeurs basses conservent plus de grain/texture (ex : 0,5–0,7), des valeurs hautes accentuent le nettoyage (ex : 0,8–1,0). Si aucun contrôle n'est visible, utilisez les réglages par défaut et ajustez la résolution ou l'encodage pour équilibrer netteté et propreté.
Cela dépend de votre GPU. Des clips 720p–1080p de quelques minutes sont un bon point de départ. Pour du long ou du 4K, traitez par segments ou réduisez la résolution, puis upscalez après restauration. Réduire la résolution ou la durée aide à éviter les limites de VRAM et accélère le traitement.
Workflows associés
- SSeedVR2
Upscale and restore detail
- SSeedVR2
Upscale a video
- Nano Banana Pro
Sharpen a product photo
- Topaz
Upscale an illustration
- HHitPaw
Restore portrait detail
- LTX-2.3
Restore archival footage
- ZZ-Image-Turbo
2.3 Creator : Upscaling et Analyse d’Image
- ZZ-Image-Turbo
Agrandir l'image : Z-Image-Turbo 2K
- Topaz
Agrandir vidéo : Topaz Starlight Precise 2.5
- NNœud partenaire
Agrandissement d'image : interpolation traditionnelle
- NNœud partenaire
Agrandissement d’Image
- NNœud partenaire
AI on the Lot 2026 - QwenVL - Améliorateur de Prompt