IMG to VIDEO simple workflow LTXV | FLORENCE | INTERPOLATION | UPSCALE — v1.1

Use this LTXV workflow to turn an input image into video; it can include Florence, frame interpolation, and upscaling, and uses listed LTXV model files, a VAE, text and vision models, and rife47.pth.

The named stages are image-to-video generation, Florence, interpolation, and upscale.

Version v1.1 adds Teacache, Torchcompile, and a GGUF version.

The input is an image, and the named output is video.

The video output is named with interpolation and upscale stages.

Listed model inputs include an LTXV checkpoint, a VAE, a text encoder, a vision model, and rife47.pth.

For the base route, place either ltxv-2b-0.9.6-dev-04-25.safetensors (7Gb) or ltxv-13b-0.9.7-dev.safetensors (29Gb) in models/checkpoints.

For the GGUF route, place ltxv-13b-0.9.7-dev-QX_0.gguf in models/unet and ltxv-13b-0.9.7-vae-BF16.safetensors in models/vae.

Custom node requirements include ComfyUI-Custom-Scripts, ComfyUI-KJNodes, ComfyUI-VideoHelperSuite, ComfyUI-mxToolkit, rgthree-comfy, and ComfyUI-Frame-Interpolation.

They also include WAS Node Suite, ComfyUI-Florence2, ComfyUI-Upscaler-Tensorrt, ComfyUI-GGUF, and ComfyUI-TeaCache.

Named model files include ltxv-2b-0.9.6-dev-04-25.safetensors, ltxv-13b-0.9.7-dev.safetensors, ltxv-13b-0.9.7-dev-QX_0.gguf, ltxv-13b-0.9.7-vae-BF16.safetensors, t5-v1_1-xxl-encoder-Q3_K_L.gguf, t5xxl_fp16.safetensors, rife47.pth, and MiaoshouAI/Florence-2-base-PromptGen-v2.0.

Choose the base route for a checkpoint in models/checkpoints, or the GGUF route for a quant model in models/unet together with the VAE in models/vae.

Sugestão · não verificado

Verify which GGUF filename you have: the setup lists ltxv-13b-0.9.7-dev-QX_0.gguf, while another listed filename is ltxv-13b-0.9.7-dev-Q8_0.gguf.

Sugestão · não verificado

Before setup, verify the exact names of ltxv-13b-0.9.7-vae-BF16.safetensors, comfyui-florence2, was-node-suite-comfyui, ComfyUI-GGUF, and teacache.

Sugestão · não verificado