AI on the Lot 2026 - WF Final - t2v
Le sixième et dernier workflow publié pour AI on the Lot 2026, mis à disposition des participants à l'atelier ComfyUI 101 pour ouvrir et suivre s'ils le souhaitent. Un workflow texte-vers-vidéo qui regroupe tout ce qui a été construit et utilisé dans les workflows précédents en une configuration complète.
À propos
AI on the Lot 2026 - Final WF - t2v est une chaîne complète text-to-video qui assemble la conception d’invite, la synthèse d’images et l’assemblage vidéo dans ComfyUI. Elle utilise AILab_QwenVL et AILab_QwenVL_PromptEnhancer pour transformer une idée simple en invite riche et structurée. Ensuite, le workflow génère des keyframes de haute qualité via l’étape de synthèse d’images (ID de nœud e9d2dda8-2509-4d8d-9702-a9bf00c26650), prévisualise avec PreviewAny, puis construit un clip cohérent dans l’étape vidéo (ID de nœud 99c27236-513e-427c-9ed4-49406365f0d3). SaveImage et SaveVideo fournissent des exports propres et reproductibles à chaque étape.
Techniquement, le flux est le suivant : votre invite texte de base (et éventuellement une image de référence) passe dans AILab_QwenVL pour extraire et enrichir les détails de scène. AILab_QwenVL_PromptEnhancer affine le texte en une invite compatible modèle, avec des sujets, styles et descriptions de plans plus clairs. Cette invite alimente le nœud de génération d’images (e9d2dda8-2509-4d8d-9702-a9bf00c26650) afin de produire un ou plusieurs keyframes contrôlés par seed à la résolution choisie. Ces frames sont ensuite envoyées au nœud vidéo (99c27236-513e-427c-9ed4-49406365f0d3), qui synthétise un clip selon la durée et les réglages de mouvement/consistance. PreviewAny permet d’itérer vite, tandis que SaveImage et SaveVideo figent des images fixes ou une vidéo finale une fois le résultat validé. Comme les étapes sont modulaires, vous pouvez l’utiliser en text-to-image uniquement, en image-to-video uniquement, ou en text-to-video de bout en bout avec le même graphe.
Ce que vous fournissez
- Lecture du workflow
S’exécute dans ComfyUI. Ce workflow ne peut pas s’exécuter sur cette page. Ouvrez-le dans Comfy Cloud, avec tout déjà chargé, ou téléchargez-le pour votre propre ComfyUI.
Essayer dans CloudDans le workflow
Un aperçu en lecture seule du graphe de nœuds. Ouvrez-le dans Comfy Cloud ou téléchargez-le pour modifier les nœuds.
Loading the graph
Open full-size workflow previewFichiers de ce workflow
- Auteur
- Julien Durand | MJM
Nécessite des nœuds personnalisés
- Lecture du workflow
Utiliser ce workflow depuis du code
Généré à partir du graphe et des entrées du workflow lorsque cela est possible ; sinon, les exemples enregistrés sont affichés. Vérifiez les identifiants et les points de terminaison avant d’exécuter le code, ce qui peut entraîner des frais.
What you need
- Runs on
- Comfy Cloud
- Clé
- COMFY_API_KEY
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({
"74": {
"class_type": "SaveImage",
"inputs": {
"filename_prefix": "z-image-turbo",
"images": [
"76:8",
0
]
},
"_meta": {
"title": "Save Image"
}
},
"77": {
"class_type": "AILab_QwenVL_PromptEnhancer",
"inputs": {
"attention_mode": "auto",
"custom_system_prompt": "You are an expert prompt writer for a photorealistic image generator. Given a short concept, expand it into a detailed prompt covering subject, lighting, setting, lens, and mood. One dense paragraph. No bullet points.",
"device": "auto",
"enhancement_style": "📝 Enhance",
"keep_model_loaded": True,
"max_tokens": 256,
"model_name": "Qwen3-VL-8B-Instruct",
"prompt_text": "a boxer alone in the locker room after a fight",
"quantization": "None (FP16)",
"repetition_penalty": 1.1,
"seed": 1,
"temperature": 0.7,
"top_p": 0.9,
"use_torch_compile": False
},
"_meta": {
"title": "QwenVL Prompt Enhancer"
}
},
"78": {
"class_type": "PreviewAny",
"inputs": {
"source": [
"77",
0
]
},
"_meta": {
"title": "Preview as Text"
}
},
"324": {
"class_type": "SaveVideo",
"inputs": {
"filename_prefix": "video/LTX_2.3_i2v",
"format": "auto",
"format.codec": "auto",
"video": [
"326:310",
0
]
},
"_meta": {
"title": "Save Video"
}
},
"330": {
"class_type": "AILab_QwenVL",
"inputs": {
"attention_mode": "auto",
"custom_prompt": "You are writing a motion prompt for an image-to-video generator that creates 5-second clips with synced audio. Describe the motion that unfolds over the next 5 seconds. State subject motion (or stillness) and camera motion as separate sentences. Be explicit about what does NOT move to prevent hallucinations. Include environmental motion (smoke, fog, light, water, fabric). Optional ambient audio at the end. Do not redescribe static visual content. One paragraph focused on motion.",
"keep_model_loaded": True,
"max_tokens": 512,
"model_name": "Qwen3-VL-8B-Instruct",
"preset_prompt": "🎬 Cinematic Description",
"quantization": "None (FP16)",
"seed": 1
},
"_meta": {
"title": "QwenVL"
}
},
"331": {
"class_type": "PreviewAny",
"inputs": {
"source": [
"330",
0
]
},
"_meta": {
"title": "Preview as Text"
}
},
"326:276": {
"class_type": "RandomNoise",
"inputs": {
"noise_seed": 42
},
"_meta": {
"title": "RandomNoise"
}
},
"326:277": {
"class_type": "RandomNoise",
"inputs": {
"noise_seed": 1072283399870947
},
"_meta": {
"title": "RandomNoise"
}
},
"326:278": {
"class_type": "LTXVConcatAVLatent",
"inputs": {
"audio_latent": [
"326:307",
1
],
"video_latent": [
"326:288",
0
]
},
"_meta": {
"title": "Concat AV Latent"
}
},
"326:279": {
"class_type": "LTXVAudioVAELoader",
"inputs": {
"ckpt_name": "ltx-2.3-22b-distilled-fp8.safetensors"
},
"_meta": {
"title": "Load LTXV Audio VAE"
}
},
"326:280": {
"class_type": "KSamplerSelect",
"inputs": {
"sampler_name": "euler_cfg_pp"
},
"_meta": {
"title": "KSamplerSelect"
}
},
"326:281": {
"class_type": "ManualSigmas",
"inputs": {
"sigmas": "0.85, 0.7250, 0.4219, 0.0"
},
"_meta": {
"title": "ManualSigmas"
}
},
"326:282": {
"class_type": "CFGGuider",
"inputs": {
"cfg": 1,
"model": [
"326:285",
0
],
"negative": [
"326:284",
1
],
"positive": [
"326:284",
0
]
},
"_meta": {
"title": "CFG Guider"
}
},
"326:283": {
"class_type": "SamplerCustomAdvanced",
"inputs": {
"guider": [
"326:314",
0
],
"latent_image": [
"326:318",
0
],
"noise": [
"326:277",
0
],
"sampler": [
"326:291",
0
],
"sigmas": [
"326:306",
0
]
},
"_meta": {
"title": "SamplerCustomAdvanced"
}
},
"326:284": {
"class_type": "LTXVCropGuides",
"inputs": {
"latent": [
"326:307",
0
],
"negative": [
"326:304",
1
],
"positive": [
"326:304",
0
]
},
"_meta": {
"title": "LTXVCropGuides"
}
},
"326:285": {
"class_type": "LoraLoaderModelOnly",
"inputs": {
"lora_name": "ltx-2.3-22b-distilled-lora-384-1.1.safetensors",
"model": [
"326:316",
0
],
"strength_model": 0.5
},
"_meta": {
"title": "Load LoRA"
}
},
"326:286": {
"class_type": "ResizeImagesByLongerEdge",
"inputs": {
"images": [
"326:290",
0
],
"longer_edge": 1536
},
"_meta": {
"title": "Resize Images by Longer Edge (DEPRECATED)"
}
},
"326:287": {
"class_type": "LTXVLatentUpsampler",
"inputs": {
"samples": [
"326:307",
0
],
"upscale_model": [
"326:311",
0
],
"vae": [
"326:316",
2
]
},
"_meta": {
"title": "LTXVLatentUpsampler"
}
},
"326:288": {
"class_type": "LTXVImgToVideoInplace",
"inputs": {
"bypass": [
"326:302",
0
],
"image": [
"326:289",
0
],
"latent": [
"326:287",
0
],
"strength": 1,
"vae": [
"326:316",
2
]
},
"_meta": {
"title": "LTXVImgToVideoInplace"
}
},
"326:289": {
"class_type": "LTXVPreprocess",
"inputs": {
"image": [
"326:286",
0
],
"img_compression": 18
},
"_meta": {
"title": "LTXV Preprocess"
}
},
"326:290": {
"class_type": "ResizeImageMaskNode",
"inputs": {
"input": [
"76:8",
0
],
"resize_type": "scale dimensions",
"resize_type.crop": "center",
"resize_type.height": [
"326:299",
0
],
"resize_type.width": [
"326:312",
0
],
"scale_method": "lanczos"
},
"_meta": {
"title": "Resize Image/Mask"
}
},
"326:291": {
"class_type": "KSamplerSelect",
"inputs": {
"sampler_name": "euler_ancestral_cfg_pp"
},
"_meta": {
"title": "KSamplerSelect"
}
},
"326:292": {
"class_type": "ComfyMathExpression",
"inputs": {
"expression": "a/2",
"values.a": [
"326:312",
0
]
},
"_meta": {
"title": "Math Expression"
}
},
"326:294": {
"class_type": "ComfyMathExpression",
"inputs": {
"expression": "a/2",
"values.a": [
"326:299",
0
]
},
"_meta": {
"title": "Math Expression"
}
},
"326:295": {
"class_type": "EmptyLTXVLatentVideo",
"inputs": {
"batch_size": 1,
"height": [
"326:294",
1
],
"length": [
"326:323",
1
],
"width": [
"326:292",
1
]
},
"_meta": {
"title": "Empty LTXV Latent Video"
}
},
"326:296": {
"class_type": "LTXVImgToVideoInplace",
"inputs": {
"bypass": [
"326:302",
0
],
"image": [
"326:289",
0
],
"latent": [
"326:295",
0
],
"strength": 0.7,
"vae": [
"326:316",
2
]
},
"_meta": {
"title": "LTXVImgToVideoInplace"
}
},
"326:297": {
"class_type": "LTXVAudioVAEDecode",
"inputs": {
"audio_vae": [
"326:279",
0
],
"samples": [
"326:309",
1
]
},
"_meta": {
"title": "LTXV Audio VAE Decode"
}
},
"326:298": {
"class_type": "ComfyMathExpression",
"inputs": {
"expression": "a",
"values.a": [
"326:300",
0
]
},
"_meta": {
"title": "Math Expression"
}
},
"326:299": {
"class_type": "PrimitiveInt",
"inputs": {
"value": 1080
},
"_meta": {
"title": "Height"
}
},
"326:300": {
"class_type": "PrimitiveInt",
"inputs": {
"value": 25
},
"_meta": {
"title": "Frame Rate"
}
},
"326:301": {
"class_type": "PrimitiveInt",
"inputs": {
"value": 5
},
"_meta": {
"title": "Duration"
}
},
"326:302": {
"class_type": "PrimitiveBoolean",
"inputs": {
"value": False
},
"_meta": {
"title": "Switch to Text to Video?"
}
},
"326:303": {
"class_type": "CLIPTextEncode",
"inputs": {
"clip": [
"326:317",
0
],
"text": [
"326:319",
0
]
},
"_meta": {
"title": "CLIP Text Encode (Prompt)"
}
},
"326:304": {
"class_type": "LTXVConditioning",
"inputs": {
"frame_rate": [
"326:298",
0
],
"negative": [
"326:313",
0
],
"positive": [
"326:303",
0
]
},
"_meta": {
"title": "LTXVConditioning"
}
},
"326:305": {
"class_type": "LTXVEmptyLatentAudio",
"inputs": {
"audio_vae": [
"326:279",
0
],
"batch_size": 1,
"frame_rate": [
"326:298",
1
],
"frames_number": [
"326:323",
1
]
},
"_meta": {
"title": "Empty LTXV Latent Audio"
}
},
"326:306": {
"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"
}
},
"326:307": {
"class_type": "LTXVSeparateAVLatent",
"inputs": {
"av_latent": [
"326:283",
0
]
},
"_meta": {
"title": "Separate AV Latent"
}
},
"326:308": {
"class_type": "SamplerCustomAdvanced",
"inputs": {
"guider": [
"326:282",
0
],
"latent_image": [
"326:278",
0
],
"noise": [
"326:276",
0
],
"sampler": [
"326:280",
0
],
"sigmas": [
"326:281",
0
]
},
"_meta": {
"title": "SamplerCustomAdvanced"
}
},
"326:309": {
"class_type": "LTXVSeparateAVLatent",
"inputs": {
"av_latent": [
"326:308",
0
]
},
"_meta": {
"title": "Separate AV Latent"
}
},
"326:310": {
"class_type": "CreateVideo",
"inputs": {
"audio": [
"326:297",
0
],
"bit_depth": "auto",
"codec": "none",
"color_space": "sRGB",
"fps": [
"326:298",
0
],
"images": [
"326:315",
0
]
},
"_meta": {
"title": "Create Video"
}
},
"326:311": {
"class_type": "LatentUpscaleModelLoader",
"inputs": {
"model_name": "ltx-2.3-spatial-upscaler-x2-1.1.safetensors"
},
"_meta": {
"title": "Load Latent Upscale Model"
}
},
"326:312": {
"class_type": "PrimitiveInt",
"inputs": {
"value": 1920
},
"_meta": {
"title": "Width"
}
},
"326:313": {
"class_type": "CLIPTextEncode",
"inputs": {
"clip": [
"326:317",
0
],
"text": "pc game, console game, video game, cartoon, childish, ugly"
},
"_meta": {
"title": "CLIP Text Encode (Prompt)"
}
},
"326:314": {
"class_type": "CFGGuider",
"inputs": {
"cfg": 1,
"model": [
"326:285",
0
],
"negative": [
"326:304",
1
],
"positive": [
"326:304",
0
]
},
"_meta": {
"title": "CFG Guider"
}
},
"326:315": {
"class_type": "VAEDecodeTiled",
"inputs": {
"overlap": 64,
"samples": [
"326:309",
0
],
"temporal_overlap": 4,
"temporal_size": 4096,
"tile_size": 768,
"vae": [
"326:316",
2
]
},
"_meta": {
"title": "VAE Decode (Tiled)"
}
},
"326:316": {
"class_type": "CheckpointLoaderSimple",
"inputs": {
"ckpt_name": "ltx-2.3-22b-distilled-fp8.safetensors"
},
"_meta": {
"title": "Load Checkpoint"
}
},
"326:317": {
"class_type": "LTXAVTextEncoderLoader",
"inputs": {
"ckpt_name": "ltx-2.3-22b-distilled-fp8.safetensors",
"device": "default",
"text_encoder": "gemma_3_12B_it_fp4_mixed.safetensors"
},
"_meta": {
"title": "Load LTXV Audio Text Encoder"
}
},
"326:318": {
"class_type": "LTXVConcatAVLatent",
"inputs": {
"audio_latent": [
"326:305",
0
],
"video_latent": [
"326:296",
0
]
},
"_meta": {
"title": "Concat AV Latent"
}
},
"326:319": {
"class_type": "PrimitiveStringMultiline",
"inputs": {
"value": [
"330",
0
]
},
"_meta": {
"title": "Prompt"
}
},
"326:323": {
"class_type": "ComfyMathExpression",
"inputs": {
"expression": "a * b + 1",
"values.a": [
"326:301",
0
],
"values.b": [
"326:300",
0
]
},
"_meta": {
"title": "Math Expression (length)"
}
},
"76:11": {
"class_type": "ModelSamplingAuraFlow",
"inputs": {
"model": [
"76:28",
0
],
"sampling": "flow",
"shift": 3
},
"_meta": {
"title": "ModelSamplingAuraFlow"
}
},
"76:13": {
"class_type": "EmptySD3LatentImage",
"inputs": {
"batch_size": 1,
"height": 512,
"width": 912
},
"_meta": {
"title": "EmptySD3LatentImage"
}
},
"76:27": {
"class_type": "CLIPTextEncode",
"inputs": {
"clip": [
"76:30",
0
],
"text": [
"77",
0
]
},
"_meta": {
"title": "CLIP Text Encode (Prompt)"
}
},
"76:28": {
"class_type": "UNETLoader",
"inputs": {
"unet_name": "z_image_turbo_bf16.safetensors",
"weight_dtype": "default"
},
"_meta": {
"title": "Load Diffusion Model"
}
},
"76:29": {
"class_type": "VAELoader",
"inputs": {
"vae_name": "ae.safetensors"
},
"_meta": {
"title": "Load VAE"
}
},
"76:3": {
"class_type": "KSampler",
"inputs": {
"cfg": 1,
"denoise": 1,
"latent_image": [
"76:13",
0
],
"model": [
"76:11",
0
],
"negative": [
"76:33",
0
],
"positive": [
"76:27",
0
],
"sampler_name": "res_multistep",
"scheduler": "simple",
"seed": 65154781543635,
"steps": 8
},
"_meta": {
"title": "KSampler"
}
},
"76:30": {
"class_type": "CLIPLoader",
"inputs": {
"clip_name": "qwen_3_4b.safetensors",
"device": "default",
"type": "lumina2"
},
"_meta": {
"title": "Load CLIP"
}
},
"76:33": {
"class_type": "ConditioningZeroOut",
"inputs": {
"conditioning": [
"76:27",
0
]
},
"_meta": {
"title": "Conditioning Zero Out"
}
},
"76:337": {
"class_type": "UpscaleModelLoader",
"inputs": {
"model_name": "4x-UltraSharp.pth"
},
"_meta": {
"title": "Load Upscale Model"
}
},
"76:63": {
"class_type": "VAEDecode",
"inputs": {
"samples": [
"76:3",
0
],
"vae": [
"76:29",
0
]
},
"_meta": {
"title": "VAE Decode"
}
},
"76:70": {
"class_type": "FL_UpscaleModel",
"inputs": {
"batch_size": 1,
"downscale_by": 0.5,
"image": [
"76:63",
0
],
"precision": "32",
"rescale_method": "bicubic",
"upscale_model": [
"76:337",
0
]
},
"_meta": {
"title": "FL Upscale Model"
}
},
"76:71": {
"class_type": "FL_KsamplerPlus",
"inputs": {
"batch_size": 4,
"cfg": 1,
"denoise": 0.2,
"image": [
"76:70",
0
],
"input_type": "image",
"model": [
"76:11",
0
],
"negative": [
"76:33",
0
],
"overlap": 0.05,
"positive": [
"76:27",
0
],
"sampler_name": "res_multistep",
"scheduler": "simple",
"seed": 974,
"steps": 9,
"use_sliced_conditioning": True,
"vae": [
"76:29",
0
],
"x_slices": 1,
"y_slices": 2
},
"_meta": {
"title": "FL KSampler Plus"
}
},
"76:8": {
"class_type": "VAEDecode",
"inputs": {
"samples": [
"76:71",
3
],
"vae": [
"76:29",
0
]
},
"_meta": {
"title": "VAE Decode"
}
}
})
# run() also passes api_key to the partner nodes in this workflow.
job = client.run(workflow, api_key=api_key)Questions-réponses
Vous devez charger correctement les nœuds personnalisés AILab_QwenVL ; ils s’appuient sur des ressources Qwen‑VL pour l’amélioration d’invite. Les nœuds de synthèse d’image et de vidéo (e9d2dda8‑… et 99c27236‑…) utilisent la configuration de modèle/checkpoint fournie par le workflow ou celle que vous connectez. Si ComfyUI signale des modèles manquants, suivez les avertissements des nœuds pour télécharger ou pointer vers les bons fichiers.
Essayez d’augmenter la force de mouvement/consistance dans le nœud vidéo, de baisser la force du bruit et de garder une seed fixe à l’étape image. Des invites claires et spécifiques issues de AILab_QwenVL_PromptEnhancer (sujets, caméra, style) aident aussi. Des clips plus courts et des résolutions modérées sont plus faciles à stabiliser pendant l’itération.
Baissez la résolution, réduisez les steps dans le nœud image (e9d2dda8‑…), utilisez une durée ou un fps plus faible dans le nœud vidéo (99c27236‑…), et gardez une seed fixe. Utilisez PreviewAny pour valider la direction avant de lancer un rendu en qualité finale.
Les deux. Vous pouvez vous arrêter après l’étape image et exporter des images fixes (SaveImage), ou injecter des images existantes/de référence via les étapes d’invite et de vidéo pour produire du mouvement. Le chemin de bout en bout est du text‑to‑video, mais les étapes sont modulaires, donc vous pouvez combiner selon vos besoins.
Workflows associés
- AACE-Step
ACE-Step 1.5 Génération de Musique (4B LLM)
- NNœud partenaire
AI on the Lot 2026 - QwenVL - Améliorateur de Prompt
- NNœud partenaire
AI on the Lot 2026 - QwenVL - i2t
- Nœud partenaire
AI on the Lot 2026_LTX-2.3_IC LoRA (Union Control)
- NNœud partenaire
Arrow2: Image to SVG
- NNœud partenaire
Arrow2: Text to SVG
- Seedance 2.0
Capture de mouvement par profondeur avec Seedance 2.0
- ZZ-Image-Turbo
Changer de nœud
- NNœud partenaire
Collages en Mouvement
- Gemini3 Pro Image Preview, Google, Nano Banana Pro
Colorer un dessin au trait
- Flux.2 Dev, Flux.2 Klein, GPT-Image-1.5, Grok, Nano Banana Pro, Qwen-Image-Edit, Seedream
Comparaison de 1 entrée et de plusieurs modèles d'édition
- Nano Banana Pro, Google
Couverture de magazine et design d'emballage