Lightning Lora, massive speed up for Wan2.1 / Wan2.2 made by Lightx2v / Kijai

Lightning Lora, massive speed up for Wan2.1 / Wan2.2 made by Lightx2v / Kijai - 2.2-Lightning-I2V-1030-H

NSFW

2.2

For I2V this motion helper node is extremely useful:

https://github.com/princepainter/ComfyUI-PainterI2V

10/30 High lora was further refined.

New I2V 1022 versions are out. They have by far the best prompt following / motion quality yet. (The lora key warning is fine, it just contains extra modulation keys that comfyui does not use. It does not matter.)

https://github.com/VraethrDalkr/ComfyUI-TripleKSampler

T2V versions just got updated 09/28. Probably still best to use a step or two with cfg / without the lora to establish motion with high noise as usual like:

2 steps high noise without the low-step lora at 3.5 CFG

2 steps high noise with lora and 1 CFG

2-4 steps low noise with lora and 1 CFG

Its definitely a big improvement either way.

T2V:

Using their full 'dyno' model for your high model seems best.

"On Sep 28, 2025, we released two models, Wan2.2-T2V-A14B-4steps-lora-250928 and Wan2.2-T2V-A14B-4steps-250928-dyno. The two models share the same low-noise weight. Wan2.2-T2V-A14B-4steps-250928-dyno delivers superior motion rendering and camera response, with object movement speeds that closely match those of the base model. For projects requiring highly dynamic visuals, we strongly recommend using Wan2.2-T2V-A14B-4steps-250928-dyno. Below, you will find some showcases for reference."

https://huggingface.co/lightx2v/Wan2.2-Lightning/blob/main/Wan2.2-T2V-A14B-4steps-250928-dyno/Wan2.2-T2V-A14B-4steps-250928-dyno-high-lightx2v.safetensors

2.1

7/15 update: I added the new I2V lora, seems to have much better motion than using the old text to video lora on a image to video model. Example is 4 steps, 1 CFG, LCM sampler, 8 shift. I uploaded the new version of the T2V one also.

I'm also putting up the rank 128 versions extracted by Kijai, they are double the size but are slightly better quality.

I suggest using it with the Pusa V1 lora as well, it seems to improve movement even more: https://huggingface.co/Kijai/WanVideo_comfy/tree/main/Pusa

No need for 2 sampler WF anymore IMO. Just plug it into your normal WF with 1 CFG and 4 steps or so. Could prob sharpen it with another pass if you wanted but it no longer hurts the movement like before for image to video.

Full Image to video Lightx2V model: https://huggingface.co/lightx2v/Wan2.1-I2V-14B-480P-StepDistill-CfgDistill-Lightx2v/tree/main/distill_models

Old:
lightx2v made a 14B self forcing model that is a massive improvement compared to Causvid / Accvid. Kijai extracted it as a lora. Example above was generated in about 35 seconds on a 4090 using 4 steps, lcm, 1 cfg, 8 shift, still playing with settings to see what is best.

Please don't send me buzz or anything, give the lightx2v team or kijai support if anyone.

Description

FAQ

Comments (28)

Lora_AddictOct 30, 2025

CivitAI

I tested all this new stuff but my I2V two sampler setup using the Lightning I2V 4 Step lora AND the old lightx2v I2V lora, all in different strengths still works WAY better in terms of quality and movement.

Ada321

Author

Oct 30, 2025

Low lora sure, high lora the latest one follows prompts far better in my experience, especially for 2D / anything not realistic.

volianticq5953Nov 12, 2025

Lora_Addict, Could you please tell me what kind of lora (lora names) you use and what weight they are for High and Low . for I2V? I'm trying to figure it out now and I don't understand which ones work best.

Lora_AddictNov 14, 2025· 1 reaction

@volianticq5953

Wan2.2-Lightning_I2V-A14B-4steps-lora_HIGH_fp16 - 1.5

Wan2.2-Lightning_I2V-A14B-4steps-lora_LOW_fp16 - 1.0

lightx2v_I2V_14B_480p_cfg_step_distill_rank128_bf16 - high 2.5

lightx2v_I2V_14B_480p_cfg_step_distill_rank128_bf16 - low 0.5

JCPornysOct 31, 2025

CivitAI

LOW version doesn't work with LORAs

Ada321

Author

Oct 31, 2025

? I use it all the time.

IrakTheOracleNov 2, 2025· 4 reactions

CivitAI

why doesn't the civit team upload these necessary loras?

Baka_OppaiNov 5, 2025· 3 reactions

CivitAI

1030 I2V seems to be a lot less able (than 1022) to work with close ups and unique positions it keeps overwriting the image with random people.

rogeakocil608Dec 11, 2025· 1 reaction

CivitAI

This lora makes motions much better! But there is an issue: sometimes it may turn the video's color to green if you set high strenth, even the strength on 1.00.

rogeakocil608Dec 11, 2025

i mean the 1030 version

fronyaxDec 20, 2025

CivitAI

T2V 1227 lora is here..

https://huggingface.co/lightx2v/Wan2.2-Distill-Loras/tree/main

essence25Jan 6, 2026

What exactly does the 1227 number represent? Also, do I need all the lora's above or only the latest ones? eg H/L1022 ? I'm still learning. ty!

fronyaxJan 7, 2026· 1 reaction

@essence25 i assume it's just a version based on the date it was released, 1227 means December 27 version. 1227 is for T2V, 1022 or 1030 is the latest for I2V. Choose and pick one based on your needs. And dont use 1227 and 1030 and 1022 together lol

essence25Jan 7, 2026

@fronyax You mean use only pairs together? Eg H/L 1022 ? When the workflow specifies a high and a low. And not mix dates?

fronyaxJan 7, 2026· 4 reactions

@essence25 Should be like this, You can use :

H1030 and L1022 or

H1022 and L1022.

Pick only one high and one low on each version, that's it

Devilday666Dec 23, 2025· 1 reaction

CivitAI

Should the strength the left at 1 for high and low? I know for Wan 2.1 Lightx have been setting high at 3 and low at 1.5...

unphasedDec 26, 2025· 2 reactions

CivitAI

I'm not really sure what I'm doing but I'm just fiddling with the bog standard wan 2.2 i2v workflow provided default with comfyui and I found after doing some experiments with full wan finetunes that I can still get impressive (let's just say, my best so far) results out of combining the original wan models with the lightx2v loras and these motion loras.

I have the best results so far setting first ksampler to 4 steps and doing 2 steps, and setting the second ksampler to 6 steps and starting from step 2. Both are on euler ancestral simple. i have lightx2v at 0.9 on high and 0.6 on low. the high motion loras here seem to give a lot of creative freedom and good results no matter what intensity to add them in.

I'll also continue to experiment with finetunes and see what weight combos work well with those but i suspect each finetune will have a different amount already included so that is gonna be a lot of trial and error.

burn1one1one812Mar 3, 2026

can you tell us what gpu youre running and what your gen times/speeds are with that settup?

conkikhon406Dec 26, 2025· 3 reactions

CivitAI

i2v h1022 work better with l1022 than h1033 (sightly better prompt accuracy)

essence25Jan 6, 2026

Did you mean "h1030" ?

conkikhon406Jan 7, 2026

@essence25 yup

velantegJan 16, 2026

CivitAI

those loras causing snowfall effect artifacts on video and distortion. Was not happening with 2.1 loras.
Is it requre specific different workflow?

clzpetn804Jan 27, 2026· 10 reactions

CivitAI

You folks should really pull the Torch Compile node out of your default attached workflow. Most people aren't benchmarking on a supercomputer if they are using your LoRA. It only spends unnecessary compute and wastes time for the vast majority of people. Thanks

ricoss35852Feb 15, 2026· 3 reactions

CivitAI

Hello,

Is it normal that we can't download the

"Wan_2_2_I2V_A14B_HIGH_lightx2v_4step_lora_v1030_rank_64_bf16" in the LOW version?

srisagerh463Mar 29, 2026

I wonder the same thing??

smart136plus2655Apr 9, 2026

CivitAI

I have a problem. When I add additional LoRa, for example, for certain actions, the image deteriorates; it becomes blurry, and my face and eyes blur. I downloaded the latest version, 1030 for high and 1022 for low, but it makes no difference. It's still just as bad, even at 1280x720 resolution.

fronyaxApr 14, 2026· 3 reactions

CivitAI

They rreleased the new 260412

https://huggingface.co/obsxrver/wan2.2-i2v-lightx2v-260412/tree/main

Only4u_ArtApr 19, 2026

who is they?
looks shady as hell

LORA

Wan Video 2.2 I2V-A14B

by Ada321

Download (Beta) View on CivitAI

base model

wan