CivArchive
    LTX 2.5 Ultimate 12GB — T2V/I2V/FFLF + Pose + 12 VFX LoRAs + RelayVSR x2 - v3.2.1
    NSFW

    ☕ One coffee = one happy creator: https://buymeacoffee.com/the_frizzy1


    Every LTX 2.5 mode in one workflow, picked from one dropdown. Text to video, image to video, first+last frame, pose control, and Lightricks' 12 VFX LoRAs: remove a person, day to night, add water, alpha matte, layout to render. Video comes out with sound.

    What you need

    ✔ Custom nodes — None. Core ComfyUI only.
    ✔ ComfyUI version — 0.45.0 or newer (for the RelayVSR nodes). On 0.44 use the v3.1.1 file in source/.
    ✔ VRAM — 12 GB — tested, not estimated
    ✔ System RAM — 16–19 GB peak
    ✔ Tested on — RTX 3060 12 GB · 32 GB RAM · ComfyUI 0.45.0
    ✔ Licence — Workflow MIT · LTX 2.5 and the LoRAs: LTX-2.x Community Licence

    Models to download

    Click a file name to download it. Put it in the folder shown, inside ComfyUI/models/. Every loader in the workflow also carries its own link, so ComfyUI's missing-models panel gives you a Download button per file.

    Required — 8 files, about 39.7 GB

    diffusion_models/

    File: ltx-2.5-22b-distilled-transformer-nvfp4.safetensors — 18.7 GB (12 GB (tested))
    Download:
    https://huggingface.co/Lightricks/LTX-2.5/resolve/main/diffusion_models/ltx-2.5-22b-distilled-transformer-nvfp4.safetensors

    Put it here:
    ComfyUI/models/diffusion_models/

    File: rt_detr_v4-x-hgnet_fp16.safetensors — 0.1 GB (Pose mode)
    Download:
    https://huggingface.co/Comfy-Org/SDPose/resolve/main/diffusion_models/rt_detr_v4-x-hgnet_fp16.safetensors

    Put it here:
    ComfyUI/models/diffusion_models/

    text_encoders/

    File: gemma4-12b-with-proj-ltx-2.5-comfy-int8-convrot.safetensors — 15.4 GB (all modes)
    Download:
    https://huggingface.co/Lightricks/LTX-2.5/resolve/main/text_encoders/gemma4-12b-with-proj-ltx-2.5-comfy-int8-convrot.safetensors

    Put it here:
    ComfyUI/models/text_encoders/

    vae/

    File: ltx-2.5-video-vae-bf16.safetensors — 1.5 GB (all modes)
    Download:
    https://huggingface.co/Lightricks/LTX-2.5/resolve/main/vae/ltx-2.5-video-vae-bf16.safetensors

    Put it here:
    ComfyUI/models/vae/

    File: ltx-2.5-audio-vae-bf16.safetensors — 0.4 GB (all modes)
    Download:
    https://huggingface.co/Lightricks/LTX-2.5/resolve/main/vae/ltx-2.5-audio-vae-bf16.safetensors

    Put it here:
    ComfyUI/models/vae/

    latent_upscale_models/

    File: ltx-2.5-latent-spatial-upscaler-x2-bf16-1.0.safetensors — 1.0 GB (T2V / I2V)
    Download:
    https://huggingface.co/Lightricks/LTX-2.5/resolve/main/latent_upscale_models/ltx-2.5-latent-spatial-upscaler-x2-bf16-1.0.safetensors

    Put it here:
    ComfyUI/models/latent_upscale_models/

    loras/

    File: ltx-2.3-22b-ic-lora-union-control-ref0.5.safetensors — 0.7 GB (Pose mode)
    Download:
    https://huggingface.co/Lightricks/LTX-2.3-22b-IC-LoRA-Union-Control/resolve/main/ltx-2.3-22b-ic-lora-union-control-ref0.5.safetensors

    Put it here:
    ComfyUI/models/loras/

    checkpoints/

    File: sdpose_wholebody_fp16.safetensors — 1.9 GB (Pose mode)
    Download:
    https://huggingface.co/Comfy-Org/SDPose/resolve/main/checkpoints/sdpose_wholebody_fp16.safetensors

    Put it here:
    ComfyUI/models/checkpoints/

    The LTX 2.5 files and every LoRA need Agree and access on their Lightricks page first, logged in to Hugging Face.

    VFX LoRAs — optional, download only the ones you use. All go in loras/.

    RelayVSR x2 — only for the RelayVSR switch. Needs ComfyUI 0.45.0+.

    diffusion_models/

    File: relayvsr_x2_lightx2v_int8_convrot.safetensors — 5.16 GB
    Download:
    https://huggingface.co/Comfy-Org/RelayVSR/resolve/main/diffusion_models/relayvsr_x2_lightx2v_int8_convrot.safetensors

    Put it here:
    ComfyUI/models/diffusion_models/

    vae/

    File: relayvsr_flash_vae_x2_fp32.safetensors — 0.13 GB
    Download:
    https://huggingface.co/Comfy-Org/RelayVSR/resolve/main/vae/relayvsr_flash_vae_x2_fp32.safetensors

    Put it here:
    ComfyUI/models/vae/

    Modes

    One MODE dropdown, and each path only runs when you pick it:

    T2V · I2V (first frame) · First + last frame · Pose control (IC-LoRA Union) · VFX video to video (12 LoRAs) · Layout to Render · Ingredients · Cinemagraph · Slow motion

    Two switches change speed and size: LTX x2 upscale (on by default) and RelayVSR x2 (off by default).

    Speed on my RTX 3060

    T2V, 4 s, Megapixels 0.5, models already loaded, a new seed per run:

    • on (default) — off relayvsr x2, 960x512 result, 1:19 wall, 11.1 GB peak vram

    • off — off relayvsr x2, 480x256 result, 0:35 wall, 11.3 GB peak vram

    • off — on relayvsr x2, 960x512 result, 0:43 wall, 11.1 GB peak vram

    • on — on relayvsr x2, 1920x1024 result, 1:41 wall, 11.3 GB peak vram

    Peak VRAM barely changes with the switches: the loaded LTX model itself fills the 12 GB. The off switch buys time, not VRAM headroom.

    Every mode run once at 960x512, 4 s: T2V 1:43 · I2V 2:00 · First+last 2:32 · Pose 3:44 · Clean Plate 3:27 · Layout to Render 3:30 · Cinemagraph 1:49 · Slow motion 1:49.

    What did not work

    • Ingredients: 9 of 9 runs kept the reference sheet's panel grid instead of making one shot.

    • Decompression invents text, so don't put screen recordings through it.

    • Layout to Render needs the First frame. Without one you get grey clay back.

    • Slow motion: use speed 0.2. 0.05 was not slower.

    • Strength above 1 did not help Cinemagraph, Slow motion or Ingredients.

    Strengths I settled on: Clean Plate 1.2 and name the person if they are sitting in shot · Colorization 1.3, at 1.0 the colour is dull and fades · Day to Night 1.0 with people moving · Water 1.0 at 1280x704 · everything else 1.0.

    Get the workflow

    More videos on this stuff: https://www.youtube.com/playlist?list=PLQ3PcJNbU2Ls

    Description

    v3.2.1: In-graph notes only, the graph is the same as v3.2.0 (same nodes, links, widget values and defaults). How to use note: added what it does, the inputs, the important settings, tips, an attention line and a Tested on line (RTX 3060 on ComfyUI 0.45.0; also run on ComfyUI 0.45.3 on a 24 GB card with the int8 transformer). Downloads note moved down so the two notes do not overlap.

    FAQ

    Workflows
    LTXV 2.5

    Details

    Downloads
    65
    Platform
    CivitAI
    Platform Status
    Available
    Created
    11/15/2026
    Updated
    11/20/2026
    Deleted
    -

    Files

    The_frizzy1_ltx-2.5-ultimate-12gb_v3.2.1.json