3 ms·
Not to take away from this awesome product and brilliant idea, but we are all hackers here- how can(?) I do this for free?
by hammock 4y ago
Not to take away from this awesome product and brilliant idea, but we are all hackers here- how can(?) I do this for free?
- TheObviousOne 4y ago[dead]
- hammock 4y agoHow long does it usually take?
- TheObviousOne 4y ago8-72 hours - as beta product we do manually QA (Human in the loop) in addition to our AI-scoring mechanism, so it depends on the traffic to our site. -Due to the fact that only the computation part takes around 2.5 hours. But we have plans to reduce it to below half in hour in this month.
- operator-name 4y ago1. Collect many good quality photos of yourself, in various lighting, directions and settings. 2. Train Stable Diffusion 2.0 to learn this new embedding, this is a bit of an art rather than a science. 3. Find a nice looking prompt with your new embedding that gives you the desired result. 4. Try ControlNet, LoRa, different base models, etc, always questioning your outputs - there's something subtlety off about them... 5. Continue to fine tune your prompt, input, masking and editing ??? N. Realise that it would have been easier to just get a photograper.
- TheObviousOne 4y agoThis is nice for start. You can find colabs for that todo online here: https://colab.research.google.com/github/TheLastBen/fast-stable-diffusion/blob/main/fast_stable_diffusion_AUTOMATIC1111.ipynb https://colab.research.google.com/github/TheLastBen/fast-sta... Unfortunately, in order to get better result you need to add much more components to perform quality assurance of the photorealism. Which we perform in here: https://www.metuktakim.co.il https://www.metuktakim.co.il much more robust techniques and in-between steps. You can try it for free with coupon:777. We'll appreciate if you happy with the results in your socials, and fill the feedback-form to us to improve.