Join the Pixaroma Discord server discord.gg/gggpkVgBf3 and check the ai2play-youtube channel Check my other channels: www.youtube.com/@pixaroma www.youtube.com/@altflux
Isn't channier is better then? Its takes 30 to 40 seconds for 4xscale with a very good model on my rtx4090. I used topaz with different settings and models, for video upscaling and it didn't do very good job of it. So it taking 2 mins or so for an image (4xScale) doesn't surprise me at all. The old (black/blue) version was better with video upscaling but the new (black/green) just sucks. The only down side to chainner is that you can't preview.
depends on the upscaling, the normal uspcaling without generative model takes a few seconds like 3-4 seconds, the one that generate new things in it takes longer. Same in comfyui i tested a few models, if just upscale without creating something is very fast, but if you want to change the image then it takes time.
@@AI2Playso if want to add skin detail like pores, lines etc then it will take couple of minutes. So these models, you need to download separately or does it download automatically. I will check it out today. I would really like it if it does adds some skin details to my AI images. Yeah i don't upscale from comfy as well. I usually generate images from comfy using my workflow that generates and then fixes the face. And then i use A1111 (img to img, like from 1024 to 2048 resolution ) with my custom model to add skin details and then i use chainner to upscale (with lexica 4x model). It's a very good model for AI images, it also refines the details specially in cloths. Then i edit the image in photoshop if i need to add or fix something and then for add realistic color tones/filters.
@@zaselimgamingvideos yes the models are downloaded automatically first time you use it. I am not paid to promote it or something, I just use it for a while now and it did got better in time. I use a combination of comfyui, topaz and photoshop depends on the projects. I can get up to 4k in the comfyui with flux like I did on episode 12 on pixaroma channel
Depends on the video card I guess, is quite fast on rtx4090, compared with other upscalers. Anyway I usually only update on black friday, since then prices are lower :)
Join the Pixaroma Discord server discord.gg/gggpkVgBf3 and check the ai2play-youtube channel
Check my other channels:
www.youtube.com/@pixaroma
www.youtube.com/@altflux
Great tutorial! Really enjoyed it.
Congratulations. I'm already subscribing and liking it (here from Brazil, singer Edu Yan from Spotify).
Great tutorial. Thank you for your well considered insights.
thank you 🙂
How did you do the img to video with lipsync in the beginning of the video with the still of the character? That was pretty good.
With kling ai, you can generate a video from image and if the lips are visible the entire 5 sec video you can add your audio and it will sync the lips
@AI2Play thanks I'll check out kling.
Have you tried with batch sequencing? Does it keep consistency?
I didn't tried it
I tested too a few days ago and it does not give such a good result. It takes also a lot of time for one image.
2 minutes for a 1024px upscaled 4 times on rtx4090 so is acceptable but om lower card probably takes more time
Isn't channier is better then? Its takes 30 to 40 seconds for 4xscale with a very good model on my rtx4090. I used topaz with different settings and models, for video upscaling and it didn't do very good job of it. So it taking 2 mins or so for an image (4xScale) doesn't surprise me at all. The old (black/blue) version was better with video upscaling but the new (black/green) just sucks.
The only down side to chainner is that you can't preview.
depends on the upscaling, the normal uspcaling without generative model takes a few seconds like 3-4 seconds, the one that generate new things in it takes longer. Same in comfyui i tested a few models, if just upscale without creating something is very fast, but if you want to change the image then it takes time.
@@AI2Playso if want to add skin detail like pores, lines etc then it will take couple of minutes. So these models, you need to download separately or does it download automatically. I will check it out today. I would really like it if it does adds some skin details to my AI images.
Yeah i don't upscale from comfy as well. I usually generate images from comfy using my workflow that generates and then fixes the face. And then i use A1111 (img to img, like from 1024 to 2048 resolution ) with my custom model to add skin details and then i use chainner to upscale (with lexica 4x model). It's a very good model for AI images, it also refines the details specially in cloths. Then i edit the image in photoshop if i need to add or fix something and then for add realistic color tones/filters.
@@zaselimgamingvideos yes the models are downloaded automatically first time you use it. I am not paid to promote it or something, I just use it for a while now and it did got better in time. I use a combination of comfyui, topaz and photoshop depends on the projects. I can get up to 4k in the comfyui with flux like I did on episode 12 on pixaroma channel
👌✌
🙂
meh, topaz 7 adn 6 do the same exept generative model (but its odd and toooo slow)
Depends on the video card I guess, is quite fast on rtx4090, compared with other upscalers. Anyway I usually only update on black friday, since then prices are lower :)