Flowframes

Flowframes enables via embedded functionality and customized methods to recreate the original frame set of your videos, to encode your outputs differently, as well as manage the video length, quality, audio, and subtitle merging.

The tool is still in its early development stages and it integrates different open source technologies for introducing in its engine the newly-created video frames and their proper interpolation, plus for handling the video transitions between the original and generated frames.

The app use RIFE (Real-Time Intermediate Flow Estimation), DAIN (Depth-Aware Video Frame Interpolation), FLAVR (a fast and efficient frame interpolation technique), and others.

Getting started with Flowframes is pretty straightforward as it quickly allows you to add video content, and you can easily navigate through the app's configuration tabs. In regards to the AI model you can opt for, the different options change with the chosen implementation; for example, for the default, RIFE - CUDA/Pythorch, — RIFE 1.5 (for old generated model), RIFE 2.0 (newly generated model), RIFE 2.4 (experimental updated model), etc.

After uploading a new video, you can trim the input by start and end, choose to loop your content, change the scene detection sensitivity, or opt for two different processing styles: automatic for extracting, interpolating, and encoding or step-differentiated (running all the processes separately, for manually editing the frames later).

The tool supports MP4, WEBM, AVI, MOV, GIF, MKV, or PNG (with image sequences) with different encoding options. Also, you can choose to actively track the application's processes, output each execution in a command-line utility, add new encoding arguments, override the FFmpeg automatic frame count, or manage audio and subtitles in separate stream threads.

To conclude, the program is smart, works well, and with proper knowledge and testing, you can successfully use it. Although it is still new, the app is a promising project with a nice implementation.