ba04a87d10
Refactor and improve the sag node.
...
Moved all the sag related code to comfy_extras/nodes_sag.py
2023-12-13 16:11:26 -05:00
6761233e9d
Implement Self-Attention Guidance ( #2201 )
...
* First SAG test
* need to put extra options on the model instead of patcher
* no errors and results seem not-broken
* Use @ashen-uncensored formula, which works better!!!
* Fix a crash when using weird resolutions. Remove an unnecessary UNet call
* Improve comments, optimize memory in blur routine
* SAG works with sampler_cfg_function
2023-12-13 15:52:11 -05:00
9ac0b487ac
Make --gpu-only put intermediate values in GPU memory instead of cpu.
2023-12-08 02:35:45 -05:00
cdff081023
Fix hypertile.
2023-12-07 15:22:35 -05:00
03eadbb53c
Make HyperTile deterministic
2023-12-06 21:17:56 +02:00
57d7f4464f
Add SDTurboScheduler node.
2023-11-28 13:35:32 -05:00
f30b992b18
.sigma and .timestep now return tensors on the same device as the input.
2023-11-27 16:41:33 -05:00
488de0b4df
ModelSamplingDiscreteLCM -> ModelSamplingDiscreteDistilled
2023-11-27 16:32:03 -05:00
e020ab61f9
Fix output APNG not working with ffmpeg.
2023-11-24 18:24:19 -05:00
8ad5d494d5
Fix APNG not working in ffmpeg.
2023-11-24 18:14:17 -05:00
916e9c998c
Use same default fps as webp node.
2023-11-24 11:19:23 -05:00
eff24ea6aa
Add a node to save animated PNG files. These work in ffpmeg unlike webp.
2023-11-24 11:12:10 -05:00
02ffbb2de3
Fix typo.
2023-11-23 23:20:07 -05:00
42dfae6331
Nodes to properly use the SDV img2vid checkpoint.
...
The img2vid model is conditioned on clip vision output only which means
there's no CLIP model which is why I added a ImageOnlyCheckpointLoader to
load it. Note that the unClipCheckpointLoader can also load it because it
also has a CLIP_VISION output.
SDV_img2vid_Conditioning is the node used to pass the right conditioning
to the img2vid model.
VideoLinearCFGGuidance applies a linearly decreasing CFG scale to each
video frame from the cfg set in the sampler node to min_cfg.
SDV_img2vid_Conditioning can be found in conditioning->video_models
ImageOnlyCheckpointLoader can be found in loaders->video_models
VideoLinearCFGGuidance can be found in sampling->video_models
2023-11-23 19:48:49 -05:00
871cc20e13
Support SVD img2vid model.
2023-11-23 19:41:33 -05:00
022033a0e7
Fix SaveAnimatedWEBP not working when metadata is disabled.
2023-11-23 15:39:35 -05:00
a657f96c5c
Add a node to save animated webp.
2023-11-23 14:28:41 -05:00
c3ae99a749
Allow controlling downscale and upscale methods in PatchModelAddDownscale.
2023-11-22 03:23:16 -05:00
31c5ea7b2c
Add LatentInterpolate to interpolate between latents.
2023-11-20 03:55:51 -05:00
dba4f3b4fc
Add a RepeatImageBatch node.
2023-11-19 06:09:01 -05:00
d9d8702d8d
percent_to_sigma now returns a float instead of a tensor.
2023-11-18 23:20:29 -05:00
8a451234b3
Add ImageCrop node.
2023-11-18 04:44:17 -05:00
7e3fe3ad28
Make deep shrink behave like it should.
2023-11-16 15:26:28 -05:00
bd07ad1861
Add PatchModelAddDownscale (Kohya Deep Shrink) node.
...
By adding a downscale to the unet in the first timesteps this node lets
you generate images at higher resolutions with less consistency issues.
2023-11-16 13:25:46 -05:00
dcec1047e6
Invert the start and end percentages in the code.
...
This doesn't affect how percentages behave in the frontend but breaks
things if you relied on them in the backend.
percent_to_sigma goes from 0 to 1.0 instead of 1.0 to 0 for less confusion.
Make percent 0 return an extremely large sigma and percent 1.0 return a
zero one to fix imprecision.
2023-11-16 04:23:44 -05:00
420beeeb05
Clean up and refactor sampler code.
...
This should make it much easier to write custom nodes with kdiffusion type
samplers.
2023-11-14 00:39:34 -05:00
8509bd58b4
Reorganize custom_sampling nodes.
2023-11-13 21:45:23 -05:00
61112c81b9
Add a node to flip the sigmas for unsampling.
2023-11-13 21:45:08 -05:00
ca2812bae0
Fix RescaleCFG for batch size > 1.
2023-11-10 22:05:25 -05:00
58d5d71a93
Working RescaleCFG node.
...
This was broken because of recent changes so I fixed it and moved it from
the experiments repo.
2023-11-10 20:52:10 -05:00
002aefa382
Support lcm models.
...
Use the "lcm" sampler to sample them, you also have to use the
ModelSamplingDiscrete node to set them as lcm models to use them properly.
2023-11-09 18:30:22 -05:00
ca71e542d2
Lower cfg step to 0.1 in sampler nodes.
2023-11-09 17:35:17 -05:00
844dbf97a7
Add: advanced->model->ModelSamplingDiscrete node.
...
This allows changing the sampling parameters of the model (eps or vpred)
or set the model to use zsnr.
2023-11-07 03:28:53 -05:00
4acfc11a80
add difference blend mode
2023-11-05 19:00:23 +01:00
ee74ef5c9e
Increase maximum batch size in LatentRebatch.
2023-11-02 13:07:41 -04:00
23c5d17837
Added Bayer dithering to Quantize node.
2023-10-31 22:22:40 +01:00
1443caf373
HyperTile node, can be found in: _for_testing->HyperTile
2023-10-21 05:16:38 -04:00
c2bb34d865
Implement updated FreeU as _for_testing->FreeU_V2 node
2023-10-18 02:06:49 -04:00
f8caa24bcc
Support hypernetwork with mish activation function and layer norm.
2023-10-17 12:08:03 -04:00
b5fa3d28d7
Fix MaskComposite
2023-10-13 13:40:53 -03:00
87097a11c3
Fix FeatherMask
2023-10-13 12:26:54 -03:00
20d3852aa1
Pull some small changes from the other repo.
2023-10-11 20:38:48 -04:00
877553843f
Add a CLIPSave node to save CLIP model weights.
2023-10-10 01:24:49 -04:00
7bb9f6b7e8
Add a VAESave node.
2023-10-09 01:42:15 -04:00
c16f5744e3
Fix SplitImageWithAlpha and JoinImageWithAlpha.
2023-10-08 15:52:10 -04:00
69a824e9a4
Move _for_testing/custom_sampling nodes to sampling/custom_sampling.
2023-10-08 03:20:35 -04:00
0e763e880f
JoinImageWithAlpha now works with any mask shape.
2023-10-04 15:54:34 -04:00
9212bea87c
Change a few things in #1578 .
2023-10-04 15:43:41 -04:00
214ca7197e
Corrected joining images with alpha (for RGBA input), and checking scaling conditions
2023-10-04 19:04:52 +02:00
585fb0475b
Adding default alpha when splitting RGB images
2023-10-04 19:04:52 +02:00
ece69bf28c
Change channel type to MASK (reduced redundancy, increased usability)
2023-10-04 19:04:52 +02:00
d06cd2805d
Added support for Porter-Duff image compositing
2023-10-04 19:04:48 +02:00
2ef459b1d4
Add VPScheduler node
2023-10-01 03:48:07 -04:00
8ab49dc0a4
DPMPP_SDE node.
2023-09-30 01:51:22 -04:00
213976f8c3
Add ExponentialScheduler and PolyexponentialScheduler nodes.
2023-09-29 09:05:30 -04:00
1c8ae9dbb2
Allow GrowMask node to work with batches (for AnimateDiff) ( #1623 )
...
* Allow mask batches
This allows LatentCompositeMasked -node to work with AnimateDiff. I tried to keep old functionality too, unsure if it's correct, but both single mask and batch of masks seems to work with this change.
* Update nodes_mask.py
2023-09-28 22:01:19 -04:00
66756de100
Add SamplerDPMPP_2M_SDE node.
2023-09-28 21:56:23 -04:00
26b7372805
Fix SplitSigmas.
2023-09-28 01:11:22 -04:00
76e0f8fc8f
Add function to split sigmas.
2023-09-28 00:40:09 -04:00
2bf051fda8
Add a basic node to generate sigmas from scheduler.
2023-09-28 00:30:45 -04:00
d234ca558a
Add missing samplers to KSamplerSelect.
2023-09-28 00:17:03 -04:00
1d7dfc07d5
Make add_noise in SamplerCustom a boolean.
2023-09-27 22:42:23 -04:00
1adcc4c3a2
Add a SamplerCustom Node.
...
This node takes a list of sigmas and a sampler object as input.
This lets people easily implement custom schedulers and samplers as nodes.
More nodes will be added to it in the future.
2023-09-27 22:21:18 -04:00
1d36dfb9fe
GrowMask now works with mask batches.
2023-09-26 02:53:57 -04:00
d76d71de3f
GrowMask can now be used with negative numbers to erode it.
2023-09-26 02:45:31 -04:00
d2cec6cdbf
Make mask functions work with batches of masks and images.
2023-09-25 16:19:37 -04:00
046b4fe0ee
Support batches of masks in mask composite nodes.
2023-09-25 16:02:21 -04:00
f00471cdc8
Do FreeU fft on CPU if the device doesn't support fft functions.
2023-09-24 18:09:44 -04:00
05e661e5ef
FreeU now works with the refiner.
2023-09-23 12:19:08 -04:00
fd93c759e2
Implement FreeU: Free Lunch in Diffusion U-Net node.
...
_for_testing->FreeU
2023-09-23 00:56:09 -04:00
29ccf9f471
Fix typo.
2023-09-22 01:33:46 -04:00
422d16c027
Add some nodes to add, subtract and multiply latents.
2023-09-21 22:23:01 -04:00
2b6b178173
Added support for lanczos scaling
2023-09-19 10:40:38 +02:00
0665749b1a
Move ModelSubtract and ModelAdd to advanced/model_merging
2023-09-17 02:10:06 -04:00
30de95e4b4
Add some nodes to subtract and add model weights.
2023-09-13 01:10:31 -04:00
8be46438be
Support DiffBIR SwinIR models.
2023-09-07 03:31:43 -04:00
766c7b3815
Update upscale model code to latest Chainner model code.
...
Don't add SRFormer because the code license is incompatible with the GPL.
Remove MAT because it's unused and the license is incompatible with GPL.
2023-09-02 22:27:40 -04:00
d70b0bc43c
Use the GPU for the canny preprocessor when available.
2023-08-29 17:58:40 -04:00
81d9200e18
Add node to convert a specific colour in an image to a mask.
2023-08-29 17:55:42 -04:00
fc99fa56a9
Add node to scale image to a total amount of pixels keeping aspect.
2023-08-18 02:32:39 -04:00
18e86a4010
add a node to allow growing of masks through dilation
2023-08-16 10:57:14 -04:00
6dc02c7bac
Add a "resize_source" option to Image and Latent CompositeMasked.
2023-08-15 17:51:52 -04:00
94fceb8700
Make Blur node use the image device for processing.
2023-08-14 21:08:45 -04:00
8c730dc4a7
Add an ImageCompositeMasked node.
2023-08-12 01:02:36 -04:00
1141029a4a
Add --disable-metadata argument to disable saving metadata in files.
2023-07-28 12:31:41 -04:00
5e3ac1928a
Implement modelspec metadata in CheckpointSave for SDXL and refiner.
2023-07-25 22:02:34 -04:00
a51f33ee49
Use bigger tiles when upscaling with model and fallback on OOM.
2023-07-24 19:47:32 -04:00
91ed2815d5
Add a node to merge CLIP models.
2023-07-14 02:41:18 -04:00
bdba394290
Add a canny preprocessor node.
2023-07-13 13:26:48 -04:00
a9a4ba7574
Fix merging not working when model2 of model merge node was a merge.
2023-07-08 22:31:10 -04:00
2c9d98f3e6
CLIPTextEncodeSDXL now works when prompts are of very different sizes.
2023-07-06 23:23:54 -04:00
c02f3baeaf
Now the model merge blocks node will use the longest match.
2023-07-04 00:51:17 -04:00
d94ddd8548
Add text encode nodes to control the extra parameters in SDXL.
2023-07-03 19:11:36 -04:00
9f2986318f
Move model merging nodes to advanced and add to readme.
2023-06-30 15:21:55 -04:00
9b93b920be
Add CheckpointSave node to save checkpoints.
...
The created checkpoints contain workflow metadata that can be loaded by
dragging them on top of the UI or loading them with the "Load" button.
Checkpoints will be saved in fp16 or fp32 depending on the format ComfyUI
is using for inference on your hardware. To force fp32 use: --force-fp32
Anything that patches the model weights like merging or loras will be
saved.
The output directory is currently set to: output/checkpoints but that might
change in the future.
2023-06-26 12:22:27 -04:00
05676942b7
Add some more transformer hooks and move tomesd to comfy_extras.
...
Tomesd now uses q instead of x to decide which tokens to merge because
it seems to give better results.
2023-06-24 03:30:22 -04:00
51581dbfa9
Fix last commits causing an issue with the text encoder lora.
2023-06-20 19:44:39 -04:00
bf3f271775
Add some nodes for basic model merging.
2023-06-20 19:17:03 -04:00
8883cb0f67
Add a way to set patches that modify the attn2 output.
...
Change the transformer patches function format to be more future proof.
2023-06-18 22:58:22 -04:00
f3ac938b4a
Round the mask values for bitwise operations.
2023-05-28 00:42:53 -04:00
comfyanonymous
Rafie Walker
asagi4
matt3o
tsone
Jairo Correa
MoonRide303
Jukka Seppänen
Corey