aloola wrote:

V2: faster, should use this as default if you not encountering any problems. model V2 47+ has a bug preventing you from using performance boost off.(It means it will run trt command on every video with a different resolution.)
4.13v1lite: lite is just a name, you may think of it as 4.135 which runs faster than 4.13
ensemble: this one will cost half the performance of the normal model. I couldn't see much difference from this, just use a better model instead.

try to use model in performance order: 4.6 > 4.9 > 4.13 lite > 4.13 and decide which one you like the most

4.9 is a sweet pot for me which has low artifact and runs faster than 4.13 20%

Anyway my post #1085 also applicable for RTX3080 users, what about 4.4 and 4.6 models?
They have pattern match and random zigzag artefacts, where too many same lines onscreen or objects, interpolated by error too, regardless of the overall frame!
Model 4.9 almost eliminates that effect, even with very low frame rate videos, but more GPU sensitive...

scb wrote:

So what about 4.13v2 lite, how does that rank?

IDK

scb wrote:

Could anyone fully clarify the difference between the regular models, the V2 models and lite models please?

e.g. 4.13 vs 4.13 lite vs 4.13v2 vs 4.13v2 lite.

It would help solidify my understanding!

V2 models faster

The RTX 4070 TensorRT performance (TRT boost enabled)
Model 4.4 (720p) singlethread 24>144 6x stable
Model 4.9 (720p) singlethread 24>120 5x stable
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
RIFE v2 models >
Model 4.4 (720p) singlethread 24>168 7x stable
Model 4.9 (720p) singlethread 24>144 6x stable
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Generally v2 models are 20% faster without noticeable intermediate frames quality drop, this peformance numbers achieved by single GPU thread, so you can go even higher, CPU utilization increase massively!
I dont recommend using more than 2 threads, but you can get upto 4x for 1080p videos easy!

What happens when someone upload onnx packages for 4070 users? Will work for anyone who has the same GPU or not? I have v1/v2 prebuilt for HD/FHD sources with TRT boost enabled!
What about reverse engineering method for 4060Ti and 4060 owners too?
Average onnx build time v1 ~1/1.5 min HD/FHD
Average onnx build time v2 ~2/2.5 min HD/FHD