eca7c528f4
Set 2d perspective flip to false by default
...
not to add new params to the settings for v0.4
2022-09-27 18:21:01 +03:00
f42d810a87
Default border mode: wrap -> replicate
2022-09-26 02:01:34 +03:00
50dd83c3ae
Speed issue debug. Step files now end in 05.png
2022-09-25 12:41:14 -05:00
7a637a1248
overlay_mask can be True when use_mask is True, but only overlays when use_mask is True
2022-09-25 11:24:54 -05:00
4dce339db7
Merge pull request #74 from enzymezoo-code/dev
...
Improved masking
2022-09-25 01:01:12 -05:00
7c44d668d3
Mask video with video init
2022-09-25 00:56:48 -05:00
8192b78ca8
Added option to overlay masked image after the generation. This is so the masked image does not get degraded by encoding and decoding repeatedly
2022-09-24 23:18:32 -05:00
c80d654749
Merge pull request #73 from enzymezoo-code/dev
...
Changed callback function to an object. Added printing/saving of inte…
2022-09-24 10:53:46 -05:00
abe26f4fdd
Changed callback function to an object. Added printing/saving of intermediate steps.
2022-09-24 10:49:10 -05:00
e47464d72d
Automate checking if value is number or expr
2022-09-24 12:59:20 +03:00
0913e8b944
Evaluate schedule parameters as functions of t
...
examples and troubleshooting at https://gist.github.com/kabachuha/5e165322382848114a59d2a855f49330
2022-09-24 01:39:53 +03:00
6fbfd19f05
fix dev branch by reverting
2022-09-21 19:41:42 -07:00
93fd972fe8
minor fixes sync py and ipynb fix change ui
2022-09-21 07:37:48 -07:00
b972bb1a4a
Merge pull request #65 from kabachuha/settings
...
Add custom settings import from file
2022-09-18 10:39:26 -05:00
79eb5436c0
Merge pull request #64 from kabachuha/main
...
Add perspective flipping to 2D animation mode
2022-09-17 21:25:37 -05:00
1e4eae6e97
Update: cv2-based frame unpacking (vid2frames)
...
Updated frames unpacking by cv2 with the option to overwrite / skip reunpacking.
overwrite option added under video_input settings.
2022-09-17 21:15:54 -04:00
96278d68e3
Add custom settings import from file
...
The file format is just the same as the settings dump, so you can feed your 20220908095753_settings.txt file generated after a run
2022-09-18 02:38:18 +03:00
927876517d
Merge pull request #63 from johnnypeck/main
...
Print some useful anim_args
2022-09-17 13:16:14 -05:00
4dea8dacae
Add perspective flipping to 2D animation mode
...
That way the camera angle can be adjusted for more dramatic compositions without enabling the 3D mode
2022-09-17 01:59:00 +03:00
ab3e80d251
Print some useful anim_args
...
Print some usefule anim_args
2022-09-13 02:20:27 -04:00
2ed011e10d
Prep dev branch for further development
2022-09-11 13:48:59 -05:00
a1f1632010
Minor patch
...
- small change to allow saving depths even with init video
- garbage collect and empty cuda cache at start of batch
2022-09-11 13:45:19 -05:00
1f4fa58a40
Update main repo to grab main branch of helpers
2022-09-11 11:54:41 -05:00
92263c9a2d
Hidden parameter: strength_0_no_init bool, force set strength to zero when no init image is being used
2022-09-11 00:10:34 -05:00
945f001fcc
Merge branch 'dev' of https://github.com/deforum/stable-diffusion into dev
2022-09-10 23:18:29 -05:00
ff3557b0a2
New secret setting use_alpha_as_mask to use the alpha channel of the init image as the mask
2022-09-10 23:10:23 -05:00
b8c7598009
Depth saving improvements
...
Make things more orthogonal. Depth saving now works even with depth warping off and in 2D mode. Special handling for first frame. Big ups to ScottieFox for testing help and bug reports!
2022-09-10 17:20:07 -05:00
cf95125226
Depth saving and updates
...
- Move depth inference to new file helpers/depth.py
- Merge depth stuff into new DepthModel class instead of passing around things separately
- New option to save the depth maps
- Use same depth map for warp of both prev and next for turbo
- Small fixes and clean ups
2022-09-10 14:28:00 -05:00
7ce3316dba
Fix to sanitize, clean up some unused imports and precision handling
2022-09-05 19:59:34 -05:00
a3d4320d4a
Support for turbo with 2D anims, resume support, compat with video init
2022-09-05 10:24:48 -05:00
76a32a1620
Turbo mode! diffusion_cadence param to diffuse every N frames
...
- run diffusion every N frames and do interpolation between trailing frames warped forward through time
- improves speed and coherence of animation videos
- supports just 3D at the moment
2022-09-03 17:56:58 -05:00
299fe5ae6b
Add files via upload
2022-09-02 18:12:17 -07:00
4f8fe15ccf
v0.3
2022-09-02 18:10:41 -07:00
1c0afbf103
change version and branch
2022-09-02 18:08:38 -07:00
bed13ed034
QoL changes
2022-09-02 17:58:53 -07:00
0f39e28dcc
Merge branch 'dev' of https://github.com/deforum/stable-diffusion into dev
2022-09-02 11:21:32 -05:00
2ca25f255a
During batches it will only clear the output at multiples of 32
2022-09-02 11:21:25 -05:00
f59357af56
Merge branch 'dev' of https://github.com/pharmapsychotic/stable-diffusion into dev
2022-09-02 10:48:54 -05:00
3c37400b64
Updated pinned pip dependencies to default versions on Colab runtime
...
Also add matplotlib which AdaBins uses in utils and remove CFGDenoiser from main notebook since it has moved to helper library.
2022-09-02 10:48:51 -05:00
ca945acd8f
Merge branch 'deforum:dev' into dev
2022-09-01 21:31:13 -05:00
78cb0e00b4
.
2022-09-01 21:28:07 -05:00
3ad6c0c826
3D animation and depth warping!
...
- bring in AdaBins, MiDaS, and pytorch3d-lite
- parallel anim_frame_warp_2d and anim_frame_warp_3d for 2D and 3D frame transformations
- transform_image_3d adapted from disco and optimized: eliminate round trips to disk, eliminate extra movement of tensors between cpu and gpu, avoid unnecessary trips through PIL and back
- move the *_series variables out of global scope and into DeformAnimKeys object
2022-09-01 19:16:59 -05:00
9487b60abf
add prompt print statement
2022-09-01 09:15:47 -07:00
63d00e125f
Prepared for dev branch
2022-08-31 15:49:32 -05:00
7eb1d8d593
Brightness/contrast adjustments to mask image.
2022-08-31 15:47:12 -05:00
e5f4b8383c
Merge branch 'dev' of https://github.com/deforum/stable-diffusion into inpainting_1.0
2022-08-31 00:30:44 -05:00
04b56c84a1
Added a new ddim discretization method called fill that fixes out of bounds errors. Embracing k-diffusion sigmas for the mask schedule
2022-08-31 00:29:49 -05:00
aeab8d8a9d
v0.2
2022-08-30 20:09:49 -07:00
09d78d50bf
Some cleanup of model config and load
...
- Group all setup cells in Setup section that can be collapsed
- Merge the model settings and model loading cells
- Remove unused code
- Use os.path.join
- No need to copy the yaml file around, use in place
2022-08-30 13:42:23 -05:00
cc122ea38f
Clear display with large batches, display grids when makin them.
2022-08-30 12:28:44 -05:00
kabachuha
enzymezoo-code
enzymezoo-code
deforum
deforum
Bardia Shahrestani
Johnny Peck
pharmapsychotic
pharmapsychotic