Skip to content
WorkflowListedExternal
Download

SVI 2 PRO with First/Last Frame (Keyframes) // Video Continuation // 3-Stage Loop Automation — SVI2PRO FLF V1

Combines SVI 2 Pro motion continuity with First/Last Frame control

TaskMultiCapabilitiesImage to image · UpscalingFamilyWanVersionRelease 2726400 — WanImageToVideoSVIProFLF.jsonEstimated VRAMNot yet characterizedUnratedgrade

Runtime profile

Access
External workflow
Status
Listed
Version
WanImageToVideoSVIProFLF.json
Source
Civitai
Updated
Nov 3, 2026

What you need

Model files
7.7 GiB known subtotal

Remaining hardware or download-size estimates are not available yet.

Requirements

Models

10

Node packs

14
  • ComfyMath

    Possible

    Node pack · Registry

  • comfyui_essentials

    Found

    Node pack · Registry

  • comfyui_layerstyle

    Found

    Node pack · Registry

  • comfyui-custom-scripts

    Found

    Node pack · Registry

  • Node pack · GitHub · Well-Made/ComfyUI-Wan-SVI2Pro-FLF

  • Node pack · Registry

  • comfyui-frame-interpolation

    FoundOptional

    Node pack · Registry

  • comfyui-gguf

    Possible

    Node pack · Registry

  • Node pack · Registry

  • comfyui-videohelpersuite

    Found

    Node pack · Registry

  • Fast Groups Bypasser (rgthree)

    Not resolved

    Node pack · Unknown

  • GetNode

    Not resolved

    Node pack · Unknown

  • rgthree-comfy

    Found

    Node pack · Registry

  • SetNode

    Not resolved

    Node pack · Unknown

AI-generated commentary

AI-generated explanation based on source and configuration details. Suggestions are clearly labeled.

Purpose

  • This ComfyUI workflow extends an existing video into a generated continuation using a prompt and keyframes, combining SVI 2 Pro motion continuity with Wan 2.2 First/Last Frame control.

Flow

  • Experimental 3-Part Loop Mode runs from Start to Loop entry, repeats a loop action, then moves from Loop exit to the final frame.

  • When increment is enabled, the repeated loop action can vary by cycle.

  • For loop mode, match each prompt to its corresponding image transition.

    Suggestion · not verified

Inputs and outputs

  • For video extension, provide an existing video, a prompt, and keyframes; the described output is a generated continuation.

  • For loop mode, provide Start, Loop, and End images plus three prompts.

Setup

  • The listed release is SVI2PRO FLF V1, with Wan Video 2.2 I2V-A14B as its base model.

Dependencies

  • The workflow file names these node packs: ComfyMath, comfyui_essentials, comfyui_layerstyle, comfyui-custom-scripts, comfyui-customnodes-template, comfyui-easy-use, comfyui-frame-interpolation, comfyui-gguf, comfyui-kjnodes, comfyui-videohelpersuite, and rgthree-comfy. Required:

  • Named model files include RealESRGAN_x4plus.safetensors, rife49.pth, SVI_v2_PRO_Wan2.2-I2V-A14B_HIGH_lora_rank_128_fp16.safetensors, and SVI_v2_PRO_Wan2.2-I2V-A14B_LOW_lora_rank_128_fp16.safetensors.

  • Also listed are umt5_xxl_fp8_e4m3fn_scaled.safetensors, wan_2.1_vae.safetensors, wan2.2_i2v_lightx2v_4steps_lora_v1_high_noise.safetensors, wan2.2_i2v_lightx2v_4steps_lora_v1_low_noise.safetensors, wan22EnhancedNSFWCameraPrompt_nsfwFASTMOVEV2Q8H.gguf, and wan22EnhancedNSFWCamera

Configuration

  • In V1.1, the end_samples (Last Frame) switch enables or disables First/Last Frame control without manually reconnecting inputs; when disabled, it returns to base SVI 2 PRO behavior.

  • If an output looks washed out, try the contrast-adjustment widget described for V1.1.

    Suggestion · not verified

Cautions

  • The listed release is V1, while the Last Frame switch and contrast widget are described under V1.1; the switch also requires the latest Wan-SVI2Pro-FLF nodes. Check your release and node version before relying on these controls.

    Suggestion · not verified

Does this need to be edited?

Sign in to send an edit request.

Sources

1 source
CivitaiPrimaryVersion 2726400
civitai.com/models/2424922

Verified Nov 3

Workflow instructions

Author notes from WanImageToVideoSVIProFLF.json · selected release

Here you can independently control the number of high‑noise and low‑noise steps.

Author notes from WanImageToVideoSVIProFLF.json · selected release

Here you can control the image size in megapixels while preserving the aspect ratio. The sides are scaled in steps of 16 pixels.

Author notes from WanImageToVideoSVIProFLF.json · selected release

Here you can choose the sampler.

Author notes from WanImageToVideoSVIProFLF.json · selected release

# Wan SVI 2 Pro FLF — notes This workflow is built around Wan 2.2 + SVI 2.0 Pro and FLF (First/Last Frame) logic, with custom nodes for segment stitching and end control. ## Core idea - Use **SVI 2 Pro** for long, continuous motion across segments. - Use **FLF-style end control** to hard-lock the last frame (or block of frames) of each segment. - Use **WanCutLastSlot** to remove the last temporal slot before feeding latents into the next segment, to avoid over-constraining the very end. ## Models This should work with your current models that already handle standard SVI 2 PRO, but I can't test every single one. Start with your own model combo, and if the results aren't what you expect, use the finetuned model this setup was built around until a more universal solution appears. Here are two options with speed LoRA baked in: - **SFW**: [**High noise model**](https://civitai.com/models/2053259?modelVersionId=2376074) [**Low noise model**](https://civitai.com/models/2053259?modelVersionId=2376133) - **NSFW**: [**High noise model**](https://civitai.com/models/2053259?modelVersionId=2540892) [**Low noise model**](https://civitai.com/models/2053259?modelVersionId=2540896) ## Useful links - SVI tutorial (video): [https://www.youtube.com/watch?v=-3DVJu72VhE](https://www.youtube.com/watch?v=-3DVJu72VhE) - Stable Video Infinity main page: [https://github.com/vita-epfl/Stable-Video-Infinity](https://github.com/vita-epfl/Stable-Video-Infinity) - KJNodes for ComfyUI: [https://github.com/kijai/ComfyUI-KJNodes](https://github.com/kijai/ComfyUI-KJNodes) - Wan 2.2 SVI 2 Pro native workflow (JSON example by KJ): [https://github.com/user-attachments/files/24359648/wan22_SVI_Pro_native_example_KJ.json](https://github.com/user-attachments/files/24359648/wan22_SVI_Pro_native_example_KJ.json) - Wan fp8 I2V models for ComfyUI: [https://huggingface.co/Kijai/WanVideo_comfy_fp8_scaled/tree/main/I2V](https://huggingface.co/Kijai/WanVideo_comfy_fp8_scaled/tree/main/I2V) - Lightx2v Wan 2.2 Distill LoRAs: [https://huggingface.co/lightx2v/Wan2.2-Distill-Loras/tree/main](https://huggingface.co/lightx2v/Wan2.2-Distill-Loras/tree/main) - SVI 2.0 Pro LoRAs for Wan video: [https://huggingface.co/Kijai/WanVideo_comfy/tree/main/LoRAs/Stable-Video-Infinity/v2.0](https://huggingface.co/Kijai/WanVideo_comfy/tree/main/LoRAs/Stable-Video-Infinity/v2.0) - Wan 2.2 I2V GGUF (for GGUF runtimes): [https://huggingface.co/QuantStack/Wan2.2-I2V-A14B-GGUF/tree/main](https://huggingface.co/QuantStack/Wan2.2-I2V-A14B-GGUF/tree/main)

Download files

Model artifacts

2 artifacts

svi2PROWithFirstLastFrame_svi2proFLFV1.zip

zip · 50.0 KB · SHA-256 2fdd9c55df7c…b717 · Civitai

Download file

WanImageToVideoSVIProFLF.json

json · 374 KB · SHA-256 fd43f1ea6a4e…c90f

Download file

Source excerpts

3 excerpts

Source context: 2094 downloads · Type Workflows · Base model Wan Video 2.2 I2V-A14B

Source context: 1966 downloads · Type Workflows · Base model Wan Video 2.2 I2V-A14B

Source context: 1703 downloads · Type Workflows · Base model Wan Video 2.2 I2V-A14B