talosh / flameTimewarpML

Flame Machine Learning Timewarp. based on arXiv2020-RIFE
MIT License
34 stars 9 forks source link

0.4.5 dev003 Flame 2025.0.1 "Timing" TWML Reverse Not Matching #98

Open jeffkyleflame opened 1 month ago

jeffkyleflame commented 1 month ago

This is my first time trying TWML 0.4.5 dev003 on Flame 2025.0.1. I'm using the Timing mode with a custom keyframed timewarp. I'm finding after a number of exports and restarts that since my Timing keyframe setup goes forwards and then backwards, my renders are not matching the timing of the keyframed Timewarp. It's a little hard for me to describe, but maybe to make it a little clearer, here are my frame rangers:

1-295 are unchanged 295 - 436 are mapped to 295-325 436-600 are mapped to 325-298

This results in my Timewarp node playing normal speed, slowing down, then reversed + slowed down.

When I get my TWML render after it processes, I find that 1-295 plays normal, but when it gets to 295, it does not behave the way my keyframes are set up, and it clearly just "gets it wrong". I've run some additional tests since then and have found that when a Timing keyframe setup goes forward and then backward, it seems to corrupt the data somehow.

talosh commented 1 month ago

Would it be possible to send Flame archive with this shot for me to check?Sent from my iPhoneOn 16 Jul 2024, at 20:18, jeffkyleflame @.***> wrote: This is my first time trying TWML 0.4.5 dev003 on Flame 2025.0.1. I'm using the Timing mode with a custom keyframed timewarp. I'm finding after a number of exports and restarts that since my Timing keyframe setup goes forwards and then backwards, my renders are not matching the timing of the keyframed Timewarp. It's a little hard for me to describe, but maybe to make it a little clearer, here are my frame rangers: 1-295 are unchanged 295 - 436 are mapped to 295-325 436-600 are mapped to 325-298 This results in my Timewarp node playing normal speed, slowing down, then reversed + slowed down. When I get my TWML render after it processes, I find that 1-295 plays normal, but when it gets to 295, it does not behave the way my keyframes are set up, and it clearly just "gets it wrong". I've run some additional tests since then and have found that when a Timing keyframe setup goes forward and then backward, it seems to corrupt the data somehow.

—Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you are subscribed to this thread.Message ID: @.***>

jeffkyleflame commented 1 month ago

I can't share the exact shot but since it's reproducible, I've recreated the issue on some test footage.

https://drive.google.com/drive/folders/1aa6yWXeZyjX9e6f5y8IrDddchH-jfR3Y?usp=sharing