Home Artists Posts Import Register

Downloads

Content

Local install guide: https://github.com/Sxela/WarpFusion#local-installation-guide-for-windows-venv.

If something's not working, make a new env and grab a fresh install.bat from that repo first, ask questions later :D

Changelog:

  • add lazywarp feature to generate flow and consistency on the fly
  • refactor flow extraction code
  • add lazy raft_model init
  • add flow settings to GUI
  • move flow extraction to run diffuse cell
  • add flow check and auto extraction to video export cell
  • add clip_skip
  • add an option to load settings from the resumable run when resuming (for #infinitevibes :D)
  • add colormatching (after render) for animatediff
  • add resume run for animatediff
  • add monster qr controlnet from https://huggingface.co/monster-labs/control_v1p_sd15_qrcode_monster
  • add mask clipping and invert options for all qr controlnets
  • add mask grayscale and thresholding options for all qr controlnets
  • add qr controlnet to preview only mode
  • additional safeguard for prompt length in animatediff mode
  • by default disable flow extraction for animatediff
  • by default disable normalize_cn_weights for animatediff and enable it for other models
  • fix looped image init error thanks to #ronaldgruckus
  • make run to resume detector look at 1st frames in the render folder, not on the settings' list
  • fix animatediff softcap not working on large batches
  • make the export video cell look for a latest run if the default one is not found or is empty
  • fix colormatch stylized for animatediff in a new run

2.11.2023, v0.26.3-5:

  • fix flow folder not being created in lazywarp mode
  • fix tensor size mismatch in lazywarp mode
  • make qr grayscale optional
  • fix temp_flow error thanks to rart_digital
  • fix video export blend = linear not working thanks to tiesto7878
  • make export video 1st frame path use the same format as the rest (avoid possible loss of the 1st frame)

3.11.2023, v0.26.6:

  • fix dpm sampler error in animatediff thanks to #Lullababy#0723

11.11.2023:

  • move to L

17.01.2024:

  • move to M

LazyWarp

(or LazyFlow? who knows :D)

GUI -> warp -> lazywarp

Optical flow can be now extracted during render on a per-frame basis. Doesn't affect render times much.
to make this possible, flow extraction has been moved to the Do The Run cell, and flow settings are now mainly controlled via the GUI warp tab. The old flow cell is there to initialize some variables, make folders, and define functions and will be removed in upcoming releases.

I'd suggest enabling lazywarp by default. Only uncheck it to preprocess the whole video if you are running into memory issues during render.

The render video cell now also extracts flow if it can't find any for your video.

Monster QR Controlnet

The monster qr controlnet from https://huggingface.co/monster-labs/control_v1p_sd15_qrcode_monster seems to produce better results than the vanilla qr controlnet.
Available in v1*, control_multi, and control_multi_animatediff modes.

QR mask settings

GUI -> controlnet

QR masks are now grayscale by default (before the source was unaltered, so RGB in most cases)

qr_cn_mask_clip_high
cutoff mask's top value, darkening it

qr_cn_mask_clip_low
cutoff mask's bottom value, lightening it

qr_cn_mask_thresh
Makes a mask binary from grayscale. set a threshold, all values above which will be white, and below it - black. 0 - off.

qr_cn_mask_invert
inverts the mask.

Order of operations:
1) invert
2) threshold
3 clip

QR controlnet preview is saved after all those operations are applied.


Resume run

"Resume run" is now working in animatediff mode.

Also added an option to load settings from the resumable run, if those exist. (kudos to #infinitevibes)
You can find it here:  
Do the Run! -> load_settings_from_run

While resuming from the last run, it will now look at rendered frames, not saved settings, like before. So if the run had a settings file but 0 frames, it will not be considered the last run.
The same goes for the video export cell, it will not try rendering the latest run if it has 0 frames, look for the previous one.

Clip skip

GUI -> diffuse -> clip_skip

Added clip skip option to the GUI.
It's a slider, but going below 3 doesn't really make sense :D

Local install guide:
https://github.com/Sxela/WarpFusion/blob/main/README.md

Guides made by users:

YouTube playlist with settings:
https://www.youtube.com/watch?v=wvvcWm4Snmc&list=PL2cEnissQhlCUgjnGrdvYMwUaDkGemLGq

For tech support and other questions please join our discord server:
https://discord.gg/YrpJRgVcax

Discord is the preferred method because it is nearly impossible to provide any decent help or tech support via Patreon due to its limited text formatting and inability to add screenshots or videos to comments or DMs.
Due to the recent Patreon comments update it's impossible to reply to comments from notifications anymore, so if your comment hasn't been replied to for a while, DM me.

Files

Waves rolling onto the beach or something - AI Generated video

Animatediff x Stable WarpFusion v0.26 - run 21 Settings: https://github.com/Sxela/WarpFusion/blob/main/examples/stable_warpfusion_0.26.0(21)_settings.txt

Comments

sxela

The venv install guide is up to date. What errors are you getting in console?

Jackson Doran

Wow I like this new version a lot. Half of my project was done v15-17. Gonna do the second half with this. Maybe inconsistent but might be cool to show as an example of the evolution of stable warp fusion

Mathew Huck

Im getting this error: ModuleNotFoundError: No module named 'torchvision.transforms.functional_tensor'