Видео

@Ozstudiosio There are a few different approaches to achieve this. At first glance, it seems like all you need is primarily a consistent character model, with the background generated through prompts. Another option would be to combine the character model with a "Style Reference" image for the background. You can further refine the character using inpainting-check out the "multi-character scene Pt 1" for some inspiration on this method. Additionally, we have a third method set to launch in the coming weeks. Stay tuned!

@ thanks for concern actually the first methods not work with our pipeline since we have our characters images for series done, I think the second method maybe work I’ll check the video you mentioned me to see how can achieve that specially it’s important for us when compose characters with background have ability to controls to how the depth or how much distance between characters from camera, I know still there some limitations with Ai generation but we try hits the notch of workflow pipeline as we can, cuz if that we try or subscribe with different platforms to see which will be beneficial with our studio, Lastly thanks for for your replay and if you have something guide me for our goal in your platform I’ll appreciate that again

@@Ozstudiosio anytime! we'll keep posting videos and we'll make sure to share one about integrating consistent characters into consistent backgrounds

@@scenario_gg thanks ill be wait

@quangdinhdota2388 - great question / we haven't tried (A,B) + C ... yet What should work is this: - Train A / B / C separately - Merge A + B + C - Generate (you can generate 2 or 3 characters) - You will (!) have to inpaint a bit with the Canvas - Create 7-8 images with your 3 characters - Retrain a model on these 7-8 images You should have a tri-character model. Check this tweet: x.com/emmanuel_2m/status/1852502104524382689

@havencat9337 Please reach out to support (via the web app or support@scenario.com) and share some data, they should be able to help you out!

@yobione are you talking about the Prompt Spark feature? help.scenario.com/prompt-spark

@@scenario_gg no on your video you are adding up prompt and u stack them up one by one, i can not replicate this.

​@yobione Ah, yes-this is the old UI for the prompt box, which has since been updated. It doesn’t affect the prompt’s functionality in any way; it was merely a visual design choice (though not particularly practical in hindsight).

@yobione That was an old UI we no longer use. It was far less user-friendly compared to the current version.

This is made with Immersity AI

You can find several custom-trained models for isometric styles with a top down perspective, in the "Platform Models"

It's just like when you buy a pizza or a hamburger. GPUs ain't cheap 😰

aXerpollAnimation Absolutely. You can work with your trained model, whether it was trained using Scenario or trained separately and uploaded. Additionally, it supports both SDXL and Flux models, for single- or multi-LoRAs.

@maximumblackness, here's a video that walks you through creating a training dataset in just a few easy steps: ruclips.net/video/jF3yza5aQhc/видео.html. Training a specific character model opens up a lot of possibilities, including multi-LoRA (and multi-characters), controlnet, Scenario Live, and more. Feel free to check out the other videos or give it a try!

@Arbretortue hi/bonjour - A video is coming in a few days, explaining how to create a good training set from scratch. Character Reference (aka as IP Adapter previously) is different than img2img and should work well with a 20-25 influence.

@@scenario_gg thanks for your answer👍 i guess I didn't find the right spot for characters reference. I will look for it .

@@arbretortue, here's a video that walks you through creating a training dataset in just a few easy steps: ruclips.net/video/jF3yza5aQhc/видео.html Training a specific character model opens up a lot of possibilities, including multi-LoRA (and multi-characters), controlnet, Scenario Live, and more. Feel free to check out the other videos or give it a try!

great, i will definitely use that. Thanks.

@@scenario_gg great, I will definitely use that. Thanks.

@haljordan1575 You can absolutely do that by masking in the Canvas. The process goes like this 1) Import your image to the canvas 2) Create a new layer - to the right (it will be transparent) 3) Start a rough sketch on the transparent layer 4) Draw a mask around the sketch (selection) 5) Make sure your character model is selected, then prompt + generate 6) Adjust and re-iterate. A video for a similar workflow is coming in a few days!

@dakidokino, here's a video that walks you through creating a training dataset in just a few easy steps: ruclips.net/video/jF3yza5aQhc/видео.html. Many users bring characters they've already designed separately (often hand-drawn), but it's always possible to generate a dataset of 5-8 images as a start. Training a specific character model opens up 10x more possibilities, including multi-LoRA (and multi-characters), controlnet, Scenario Live, and more. Feel free to check out the other videos or give it a try!

@@scenario_ggaah. I see now. Thank you!!

@Yousseftarek3276 here's a video on how create training dataset in a few easy steps - ruclips.net/video/jF3yza5aQhc/видео.html. Feel free to reach out!

@@scenario_gg thanks alot , great tutorial

Indeed, but it's fixable! x.com/emmanuel_2m/status/1796743430955253987

👋 cofounder here - yes you can, and it’s insanely fun

It works for anything: characters, creatures, props, backgrounds, UI, icons..

@lailaion3035 - Scenario Live now works with any plan, starting from Creator ($15 a month or $12 if billed annually)

We currently still offer a free plan, paid plans start at $15/mo