Workflow Summary:
- Prepare source video with all necessary enhancements. Colorisation, stablisation, colour and lightening enhancement etc. I have used AVC Labs video enhancement in preparing all the videos to add an exagerated "pop" element to the results.
- In Amuse Video Generation Expert Mode, import prepared video - unstable beyond 10 minutes
- Select FRAME TO FRAME
- Select previously downloaded PERMISSIVE BEAUTY LCM
- Select previously downloaded Upscaler REALESRGAN 2X
- Load selected models
- Load chosen video
- Choose Video FPS Output - 15fps is very good and fast. 30fps is slower and slightly higher quality
- Select - 8 through to 12 is fine. Much higher than 12 and some things start to get silly
- Guidance - leave an 1. Any use of this control will slow the it/s down by 40-50%. OK to use for short videos for exagerated results, but requires longer processing times.
- Strength is the "magic sauce" and seems to fuse all the other settings together. I have had success within the range of 14 through to 45 for these projects. My default starting number is 23 and moving higher or lower from there.
- FRAME BLENDING is not available for this LCM, however with some of the other LCM's its impact is very mild - like adding an extra spice in the magic sauce, not totally necessary, but an interesting twist. The LIGHTEN frame blending appeared to add a certain "je ne sais quoi" to the output. Far easier to include the "je ne sais quoi " in the video preparation stage imho! Grok has a lovely description of them that talks about beta waves and theta curves or something.
- Enter prompt. I have only ever tested basic prompts such as male, handsome, hairy, old, young etc. Without the prompt for male, the LCM has a definate default swing towards to Japanese cult female anime!
- Select GENERATE , and prepare for a very long long wait. About 3-5 hours at best. 3 hours for a 512x512 output (4.8 it/s) or 5 hours for a 768x512 output (2.7 it/s)
- If Upscaler is not selected, processing times are improved by about 45%, but final output is significantly downgraded - may be useful for DRAFT outputs.
- Once file has been processed, use regular video editing software to recombine sound from original clip to AI. video clip. If FPS was alterred during processing, use video editing to match video timelines. I use PowerDirector 365 and PowerDirector 21 for most video editing - mainly as I have been using it since version 7 (2001)
Log in or sign up for Devpost to join the conversation.