Nodes/HappyLatent/HappyLatentVideo
ComfyUI Node

HappyLatentVideo

Arrange clips, prompts and images. Choose exactly one audio mode per clip: silent, provided audio or generated audio. Connect the model, text encoder, video VAE and audio VAE, then send VIDEO to Save Video. Optionally enable spatial and/or temporal LTX 2.5 latent upscaling before decoding. Temporal upscaling includes source-anchored refinement, preserving duration and audio at double the output FPS.

By EricFillion·Created 25 days ago·Updated about 23 hours ago· 6
HappyLatentVideo
  • model
  • clip
  • video_vae
  • audio_vae
  • upscale_model
  • temporal_upscale_model
  • video
◄project{"version": 3, "id": "happylatent-project", "title": "Garden morning", "fps": 24, "width": 1024, "height": 576, "seed": 42, "globalPrompt": "A peaceful morning in a sunlit garden. Natural colors, soft golden light, and realistic cinematic photography.", "samplingPreset": "auto", "scenes": [{"id": "scene-1", "name": "Garden path", "durationFrames": 120, "prompt": "A wide shot of a garden path surrounded by flowers. A gentle breeze moves the petals and leaves as the camera slowly glides forward. Birds chirp in the distance."}, {"id": "scene-2", "name": "Butterfly", "durationFrames": 120, "prompt": "A close-up of a butterfly resting on a flower. It slowly opens and closes its wings while the flower sways gently. The camera stays steady, with soft birdsong in the background."}], "audioSegments": [{"id": "audio-1", "kind": "generate", "startFrame": 0, "endFrame": null}]}►
◄upscale_enabledfalse►
◄upscale_modespatial►
◄spatial_upscale_model_namenone►
◄temporal_upscale_model_namenone►
◄refine_enabledfalse►
CategoryHappyLatent

Inputs (12)

NameTypeDefaultDescription
modelMODEL—
clipCLIP—
video_vaeVAE—
audio_vaeVAE—
projectSTRING{"version": 3, "id": "happylatent-project", "title": "Garden morning", "fps": 24, "width": 1024, "height": 576, "seed": 42, "globalPrompt": "A peaceful morning in a sunlit garden. Natural colors, soft golden light, and realistic cinematic photography.", "samplingPreset": "auto", "scenes": [{"id": "scene-1", "name": "Garden path", "durationFrames": 120, "prompt": "A wide shot of a garden path surrounded by flowers. A gentle breeze moves the petals and leaves as the camera slowly glides forward. Birds chirp in the distance."}, {"id": "scene-2", "name": "Butterfly", "durationFrames": 120, "prompt": "A close-up of a butterfly resting on a flower. It slowly opens and closes its wings while the flower sways gently. The camera stays steady, with soft birdsong in the background."}], "audioSegments": [{"id": "audio-1", "kind": "generate", "startFrame": 0, "endFrame": null}]}—
upscale_enabledoptBOOLEANfalseUpscale each scene's generated latents before decoding using the selected mode. Timeline dimensions and frame rate remain the base generation settings.
upscale_modeloptLATENT_UPSCALE_MODELConnect the official LTX 2.5 spatial 2x latent upscaler. Used only by spatial modes.
upscale_modeoptCOMBOspatialSpatial doubles width/height; temporal doubles FPS while preserving duration and audio; combined does both. Temporal upscaling always includes source-anchored refinement.
temporal_upscale_modeloptLATENT_UPSCALE_MODELConnect the official LTX 2.5 temporal 2x latent upscaler. Used only by temporal modes.
spatial_upscale_model_nameoptCOMBOnone1 options: none
temporal_upscale_model_nameoptCOMBOnone1 options: none
refine_enabledoptBOOLEANfalseRefine the output with the connected generation model. Spatial and temporal upscaling always require refinement.

Outputs (1)

NameTypeDescription
videoVIDEO—