Now the question is..How do I lock it to that face and generate a bunch of images from different angles to train a lora or a textual inversion on them?
Now the question is..How do I lock it to that face and generate a bunch of images from different angles to train a lora or a textual inversion on them?
- Jan 4, 2022
- 241
- 947
You must be registered to see the links
Its for game developers, here is discord:
You must be registered to see the links
- Dec 21, 2020
- 1,611
- 3,863
You can't yet. I trust all of us, and the whole world, are waiting for such feature to arrive.
- Jan 4, 2022
- 241
- 947
Try to inpaint only body number of times, choice one and send it to D-ID, make video and get good frames , done!
After you made first LORA trained on 2-3 or some small number images, you load it in txt2img and generate 100+ and pic good for second training.
View attachment friendly-sky.mp4
Good idea, but WAY too mechanical. I suppose I could do it with inpaint, but that's too much work
EDIT: Oh! Now that's a good idea. Multiple trainings in an ever growing library of renders. Somehow I thought I'd train it only once.
- Jan 4, 2022
- 241
- 947
- Jan 24, 2020
- 1,401
- 3,808
Can't you do this with Live Pose in ControlNet?
You must be registered to see the links
- Dec 21, 2020
- 1,611
- 3,863
I'd argue we've never seen a complete "fly around" of a character, and hence I firmly believe this cannot be done.
Here is the context. I bet in DAZ/Blender/HS2 you can clearly make a video showing a spinning charter while you change the environments and clothes. I trust this is what everyone wants to get from AI/SD and this is the big goal. This is truly the distilled ask.
And while I watch the Control Net videos it feels like their end results fall way short of this, although the tech is promising.
But it will probably take SD 3.0 or somesuch together with the next crop of Control Net like tech before we are there.
- Jan 24, 2020
- 1,401
- 3,808
- Dec 21, 2020
- 1,611
- 3,863
Indeed, applies to CharTurner too. Once I see a video of a finished product: a chara spinning as outfits and environments change I'll be a believer. Right now, based on my understanding of how SD works, the character is materially affected by changing the environment and their outfit.
We will probably arrive at a hybrid first - you make a model in, say, DAZ, she becomes your seed, i.e. Alice123, you spin her and generate images, then you train a model, and then you bring her back in via a tag [Alice123].
Last edited:
- Jan 24, 2020
- 1,401
- 3,808
- Jun 3, 2017
- 998
- 1,542
Has anyone played around with using 3D posing to help direct generation?
You must be registered to see the links
- Jan 24, 2020
- 1,401
- 3,808
- Nov 12, 2017
- 432
- 3,524
good work, but some of the images are bit questionable, especially the first 2, I would err on the side of caution and remove
- Jan 4, 2022
- 241
- 947
https://thef95zone.info/threads/the-elders-betrayal-solaria-ch-1-7-lumifel.137827/
This game is AI only images, check it.
This game is AI only images, check it.
- Jan 24, 2020
- 1,401
- 3,808
I'm experimenting with the gorgeous prompt by atheran from this post
https://thef95zone.info/threads/ai-art-show-us-your-ai-skill.138575/post-10194065
I have reworked it quite a bit but kept the theme intact.
https://thef95zone.info/threads/ai-art-show-us-your-ai-skill.138575/post-10194065
I have reworked it quite a bit but kept the theme intact.
Last edited:
- Jan 24, 2020
- 1,401
- 3,808
- Jan 24, 2020
- 1,401
- 3,808
