v0.0.8 #374
Replies: 10 comments 17 replies
-
|
Thank you for your hard work Carson! |
Beta Was this translation helpful? Give feedback.
-
|
Really cool to see all these new features and enhancements. Love it. |
Beta Was this translation helpful? Give feedback.
-
|
Testing batch and iteration now. I noticed right away that the History display of the batch is a bit strange. Lots of empty lines. I wonder with styles you specify: custom, batch, texture, etc. I wonder if they should be a panel that can add "style presets" just to separate the prompt context and style. Btw, this latest version crashes Blender, I might do reinstall now asap. |
Beta Was this translation helpful? Give feedback.
-
|
I tested "Dream Studio" online, apparently sometimes the generation is pretty fast, almost like a second to get the result. Sometimes it does wait a bit. It is not super clear on progress and result. I am sometimes being impatience and hitting the button a few times, maybe I am sending multiple tasks. Currently the result is named with just the SEED, which is probably okey, but could the name of files include the prompt text but truncated + the seed? Furthermore, I feel a little bit more comfortable if I could see LOG text of how much processing time will take and already took. |
Beta Was this translation helpful? Give feedback.
-
|
Would be handy for an option to include some Dream Textures settings (as prompt, steps, cfg scale, sampler...) in the rendered image. When experimenting/ tweaking settings this would be very helpfull. |
Beta Was this translation helpful? Give feedback.
-
|
Using a certain same seed (not random) with size 512 x 512, gives a complete new rendering when using 1024 x 1024....that's strange isn't it?I tried the small size to find that image I want, because its very fast, use that seed with a different, bigger size but new image! |
Beta Was this translation helpful? Give feedback.
-
|
This is to be expected. Changing pretty much any parameter besides step count after a certain point will give a different result. To increase the size of an image use the built-in AI upscaling. |
Beta Was this translation helpful? Give feedback.
-
|
Thank you for the info! |
Beta Was this translation helpful? Give feedback.
-
|
I don't understand the "inpaint: mask source - prompt". I always get the same result with this option. I do not raely understand what the "Confidence Treshhold" is |
Beta Was this translation helpful? Give feedback.
-
|
You write what you want to be masked out. So if you want to replace a mountain in the background you would enter “mountain”. The confidence threshold is how confident it has to be that a pixel is part of a mountain to mask it out. So if it’s set to 50%, it has to be at least 50% sure a pixel is part of a mountain to replace it. |
Beta Was this translation helpful? Give feedback.
-
|
Ah thank you for your answer. Do I need do make an alpha channel mask when I use "mask source - prompt" also? |
Beta Was this translation helpful? Give feedback.
-
|
No, the prompt is used to generate the mask. |
Beta Was this translation helpful? Give feedback.
-
|
Thank you. That works fine Another question ist about the Outpaint. I always get a line or a hart crossing where at the expanded edge. I there a way to prevent that?
|
Beta Was this translation helpful? Give feedback.
-
|
Try increasing the blend option. Outpainting will improve in the next version using the v1.5 inpainting-specific model. |
Beta Was this translation helpful? Give feedback.
-
|
Is also possible to load the new StableDiffusion-v2? |
Beta Was this translation helpful? Give feedback.
-
|
With #405 it will be |
Beta Was this translation helpful? Give feedback.
-
|
This means in the next version v0.0.9? |
Beta Was this translation helpful? Give feedback.
-
|
Yes |
Beta Was this translation helpful? Give feedback.
-
|
Alright, noted. When is the 405 scheduled for release? |
Beta Was this translation helpful? Give feedback.
-
|
ooooOOOOOOoooooooh!
Op di 6 dec. 2022 om 14:40 schreef Carson Katri ***@***.***>:
… Yes
—
Reply to this email directly, view it on GitHub
<#374 (reply in thread)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AFPWYYZ3LEBDO3UESBPXMLLWL463XANCNFSM6AAAAAARUOPDXM>
.
You are receiving this because you commented.Message ID:
***@***.***
com>
--
*Met vriendelijke groet/ with kind regards,*
*Freddy Peters - SjonSjine*
*M:+31(0)6 50 52 61 73*
*Veemkade 592*
*1019 BL Amsterdam*
*The Netherlands*
*https://www.sjonsjine.nl <https://www.sjonsjine.nl>*
https://linktr.ee/sjonsjine - the whole package -
|
Beta Was this translation helpful? Give feedback.
-
|
Got the answer to below question: On twitter I saw a feature that uses a depth map to generate the images. Where do I acces that? I installed the add on but see no 'depth' option to select in the advanced section.
|
Beta Was this translation helpful? Give feedback.
-
|
Oh cool, I would love to try it! Would just have to find out how to do that from the page you shared. |
Beta Was this translation helpful? Give feedback.
-
|
Oh wait I misunderstood, PR is still open means pull request is still open meaning you are open for feedback and still building on it but it does not mean open for beta testers. I had to use ChatGPT to understand the lingo. I really enjoy using dream textures. Using dept maps would truly be a game changer. Especially if you could combine it with other models, like use the depth model to make a mask and another model to turn the generation into a certain style. That would be super helpful to make a scene more coherent and would also allow people to use more open sources blender objects, as you could just use the base model and retexture it in the style of your scene. Thanks for your work! |
Beta Was this translation helpful? Give feedback.
-
|
Ha cool, I could sponsor you, great feature, never used Github before |
Beta Was this translation helpful? Give feedback.
-
|
The 2+ SD models don't currently work, correct? |
Beta Was this translation helpful? Give feedback.
-
We have to wait v0.0.9 for this feature... |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
This update includes seamless upscaling, an improved render pass, a redesigned "Source Image" panel, model switching, DreamStudio integration, and more!
Choose Your Installation
Several versions are available. Find the one that will work for you.
Windows
NVIDIA GPU
dream_textures-windows-cuda.7z- Local and cloud rendering with NVIDIA CUDAdream_textures-windows-cuda.7zarchive, select "7-Zip" > "Extract Here" to get thedream_textures-windows-cuda.zipfile.dream_textures-windows-cuda.zipfile.dream_textures-windows-dreamstudio.zip- Cloud rendering with DreamStudioAMD GPU
dream_textures-windows-dreamstudio.zip- Cloud rendering with DreamStudiomacOS
Apple Silicon (M1/M2)
dream_textures-macos-arm.zip- Local and cloud rendering for Apple Silicon Macsdream_textures-macos-arm-dreamstudio.zip- Cloud rendering with DreamStudioIntel
dream_textures-macos-intel-dreamstudio.zip- Cloud rendering with DreamStudioLinux
Installation must be completed manually. Follow the instructions under Contributing for more details.
Setup
After installing the appropriate add-on ZIP file, enable the add-on and expand it's preferences. There you will find further instructions for setup.
Local Generation
To setup local generation, you will need a model checkpoint file, such as:
sd-v1-4.ckpt, and import it in the add-on preferencesv1-5-pruned-emaonly.ckpt, and import it in the add-on preferencesDreamStudio Generation
To setup DreamStudio integration, find your API key and enter it in the add-on preferences.
What's New
Seamless Upscaling
Check "Seamless" to ensure tileable results from the AI upscaler.
Render Pass Color Management
Color Management settings are now correctly applied when using the Dream Textures render pass. This change is automatically applied, and ensures results match between manual runs of Dream Textures and automated runs via the render pass.
Advanced Presets
Choose defaults and create your own presets for advanced configuration.
File Batch and Iterations
Use a text file to run multiple prompts in a row, and set Iterations above 1 to try multiple seeds without interacting with the UI.
Init Image Redesign and Outpainting
The redesigned "Source Image" panel makes manipulating images easier to use. Various bugs regarding the "Mark Inpaint Area" brush have been resolved. Outpainting allows images to be extended beyond their original size. The "Prompt" mask source when inpainting allows you to choose a region to inpaint with a simple text prompt instead of a manual alpha-channel mask.
Asymmetric Tiling
Tile along a single axis, or both for the original behavior. This can give interesting results for images such as infinitely scrolling backgrounds, decals, and more.
DreamStudio Integration
Connect your DreamStudio account to generate in the cloud. This is optional, and separate lighter builds with the DreamStudio backend only are provided for users with incompatible hardware.
Model Switching
Import multiple models and switch between them quickly to experiment with fine-tuned and updated checkpoints.
Full Changelog
New Contributors
Full Changelog: 0.0.7...0.0.8
This discussion was created from the release v0.0.8.
Beta Was this translation helpful? Give feedback.
All reactions