Kijai/WanVideo_comfy_fp8_scaled

Please provide recipe on how to convert to fp8 scaled

#43 opened 3 days ago by

firesani

OneToAllAnimation FP8 model issue with Wan2.1-Fun-14B-InP-MPS lora

#42 opened 17 days ago by

Marceltb

DEAR KIJAI ,CAN YOU PUT THE Wan21-14B-SCAIL-preview_fp8_e4m3fn_scaled_KJ for e5m2 version please.Thanks

2

#41 opened about 1 month ago by

bsarea

RuntimeError: shape '[21, 232448, 16]' is invalid for input of size 77758464

#40 opened about 1 month ago by

Deepsh2000

RTX 3090 24 GB Wan Animate v2 and SteadyDancer fp8 scaled OOM issue

1

#39 opened about 2 months ago by

Aorora12

Differences between same diffusion models from comfy-org repo

1

#38 opened about 2 months ago by

AbsAnon

FP16 safetensors for Mocha

#37 opened 2 months ago by

thegreatest1506

OVI-FP8-Scaled

#36 opened 2 months ago by

br1an

Updated Ovi 10 second workflow.

4

#35 opened 2 months ago by

neutrosophic

Kneeling down to ask the expert to come up with Mocha's E5m2 quantification model, 30

4

#34 opened 2 months ago by

wanghaohao

ChronoEdit workflow?

👀 1

2

#33 opened 3 months ago by

Aorora12

What is MoCha?

👍 1

5

#32 opened 3 months ago by

hfmaster

Wan 2.2 Animate darkening/artifacts on last 4 frames (last latent)

3

#31 opened 3 months ago by

Hesajon

Should the new anisora fp8 models generally be compatible with existing wan loras?

1

#30 opened 3 months ago by

masfutang

Hello teacher, the Wan2_2-Animate-14B_fp8_scaled_e4m3fn_KJ_v2.safetensors you sent at noon on 2025/10/7 is usable for FP8, but the one you just sent is not usable.

4

#28 opened 3 months ago by

a0983201593

Wan2_2-Animate-14B_fp8_scaled_e4m3fn_KJ_v2.safetensors (v1 worked, this one doesnt)

2

#27 opened 3 months ago by

GJohGJ

LoRA loading errors with Wan2_2-Animate-14B_fp8_scaled_e4m3fn_KJ_v2.safetensors

4

#24 opened 3 months ago by

Hesajon

Wan 2.2 VACE workflow diffusing areas outside face mask (hair, edges)?

1

#23 opened 3 months ago by

Sajjadalgburi

WANWRAPPER COM SAGE ATTENTION NO KAGGLE E COLAB

#22 opened 4 months ago by

rbaiw

How to use the conversion script?

🚀 👍 2

#18 opened 4 months ago by

Vigilence

Text encoder.

#17 opened 5 months ago by

lhucklen

Amazingly fast on the new S2V model - thanks!!

➕ 1

#16 opened 5 months ago by

cromagnone

Thank You!

😎 👍 16

1

#15 opened 5 months ago by

MintLab

Request InfiniteTalk-Multi_fp8_scaled models

1

#14 opened 5 months ago by

ZWill

skyreels scaled versions

👍 1

1

#12 opened 5 months ago by

Rudra-ai

File Size fp8 over fp16

3

#11 opened 5 months ago by

rahul7star

What is the fixed version of fun control?

1

#10 opened 5 months ago by

WhimmY

Wan 2.2 Fun Questions

9

#9 opened 5 months ago by

Grizly

How to use fp8 mat-mul?

8

#7 opened 6 months ago by

Muawiz

What changes were needed to support this model?

1

#6 opened 6 months ago by

aw632

What is the difference between this and ComfyOrg's quantitative model?

1

#5 opened 6 months ago by

Stone2013

Rename TI2V/Wan2_1-TI2V-5B_fp8_e4m3fn_scaled_KJ.safetensors to TI2V/Wan2_2-TI2V-5B_fp8_e4m3fn_scaled_KJ.safetensors

#4 opened 6 months ago by

jtsanborn

How to use with ComfyUI native WanVideo nodes?

4

#2 opened 6 months ago by

zhilengjun

can you please GGUF versions of fp8? Not sure if that makes any sense but just save some storage and processing power.

1

#1 opened 6 months ago by

ryg81