video-extend

1# Video Extension (Pusa 2.2 — temporal flowmatching)
2 
3## Overview
4 
5Pusa extends a video temporally. It continues and lengthens an existing clip
6rather than regenerating it from scratch. It does this on the
7ComfyUI-WanVideoWrapper stack (kijai) using the WAN 2.2 T2V A14B dual
8HIGH/LOW models you already have for `wan-t2v-video`, plus the small Pusa V1
9LoRAs and a Pusa-specific sampling path: the `flowmatch_pusa` scheduler and
10the `WanVideoAddPusaNoise` node. The input clip is encoded with
11`WanVideoEncode` and injected as the first latents of the generation, which
12is what carries the existing motion and content into the continuation.
13 
14The official reference graph is kijai's
15`wanvideo_2_2_14B_Pusa_extension_example_01.json` (in
16`ComfyUI-WanVideoWrapper/example_workflows/`). This skill is built directly from
17that workflow plus the live node schemas.
18 
19> Relationship to `wan-t2v-video`: Pusa rides on the exact same WanVideoWrapper
20> stack. Same T2V A14B HIGH/LOW fp8 models, same UMT5 text encoder, same WAN
21> VAE, same block-swap/torch-compile machinery. The only new downloads are the
22> two Pusa V1 LoRAs (~1.9 GB total). Read `wan-t2v-video` first for the base
23> stack; this skill is the temporal-extension delta on top of it.
24 
25> Verification note: every node, model, LoRA filename and setting below was
26> confirmed against the live ComfyUI `/object_info` (WanVideoWrapper installed)
27> and against kijai's example workflow JSON + HF repo (June 2026). Where a value
28…