Scene Detection – Flame 2018

Scene Detection – Flame 2018


Hey everyone, Grant here for the Flame Learning Channel. In this video, we will be taking a look at the new Scene Detection workflow in the Flame 2018 products. This will allow you to bring a single clip… Analyze it… And add cuts between all the shots that make up the clip. Now to get started… Scene Detection will work with any media containing multiple shots… So you can use your own media to follow along. Here I have a 50 second segment that I have already opened as a sequence. When I scrub the positioner… You can see that I have multiple shots which make up two different scenes. Now it’s not unusual to encounter this issue… Where you have been given a flattened edit which could be a commercial or film etc… And it needs to be broken apart to work on individual shots. If you were supplied an EDL, AAF or XML… You could load that on top of the source… And break up the shots automatically. However if you have just been given a flatten source clip like this and nothing else… You could use scene detection to break the clip apart quickly… Without manually scanning through the source for one cut at a time. To get started… You need to know that scene detection is selection based. So if you have video grouped with synchronous audio… Ensure they are both selected. If you just select Video and ignore the audio segments… They obviously won’t be affected by the scene detection tools. Now you can apply scene detection in two ways. If you don’t see the SCENE DETECT button in the tool bar… Just click the thumbnail in the FX pipeline. Alternatively, just call up the contextual menu over the segment… And choose SCENE DETECT. Once you’ve chosen either method… The segment is analyzed… And the pixel analysis is drawn in the video segment. Any spikes in the analysis indicate a dramatic shift in the image… Which is normally indicative of a cut. Once the analysis is complete… The Scene Detection Threshold is applied… And temporary cuts are added at frames above the threshold. You can now scrub the time bar… Or use the UP and DOWN arrow keyboard shortcuts to navigate the sequence. Before we look deeper into the Scene Detection… I just want to point out that the Scene Detection Analysis is cached per source. This means that if you were to cancel the Scene Detection… And press the SCENE DETECTION button again… You will get the same result… but without having to redo the analysis. This also includes any threshold curve adjustments. But I must mention that this only works for the current Flame session. If you reboot the software… The analysis will be performed again. Now let’s talk about the Scene Detection Threshold. Click and drag the bottom of the video track… to stretch its height in the timeline view. This is not a necessity… but the pixel analysis and threshold curve are easier to see. When you increase the threshold… The small shifts in the pixel analysis can be ignored… And this in turn will give you less cuts in the segment… Because edits will only be applied at much higher peaks in the analysis. But when you decrease the threshold… Much smaller shifts in the pixel analysis will be considered… And this will result in more edits be applied to the segment. So depending on the type of footage you’re working with… You could adjust the threshold globally across the segment. Reset the Scene Detection with the RESET button… or adjust the threshold to its default value with CONTROL+CLICK. You can also adjust the Scene Detection Threshold on the curve for more complex image data. For instance, there might be loads of movement in the image… Such as whip pan or track. Let’s look at the end of the segment as an example. Zooming into this part of the clip… There are a lot of temporary cuts. Scrubbing over this section… You can see that this is a fast pan… and the pixel analysis went up quite a bit. With this section over the threshold… Temporary cuts are being added where they are not wanted. If you want to adjust the threshold for just this section of the segment and not globally… You can edit the curve. Enable the EDIT button. The mode switches to ADD… And you can add two keyframes into the curve. As I raise the threshold on this keyframe… You will see the cuts disappear as they are below the Scene Detect Threshold. You could add a few more keyframes… And bring the threshold back down again if you want. You can set the Edit mode to ADD, MOVE or DELETE keyframes on the curve. And as a reminder… If you wanted to reset the entire curve… You could click the RESET button. Finally… The last step is to make the cuts permanent. Click the APPLY button. The splice with the equals sign indicates that these are Match Cuts. You get this when editing a source that has one continuous timecode. As a tip… If you want to remove a Match Cut like this… Select a cut and press CONTROL+ALT+X. The two segments will combine. This is obviously not what you want with scene detection… So UNDO that operation. You’ll also notice that each segment has a large amount of handles… That span the length of the original source per segment. There is no harm leaving it like this… But you could also consolidate the handles if you want. Draw a selection of the entire sequence. Call up the Contextual menu… Choose Commit… And Consolidate Handles. Press Zero and enter. All the segments have zero handles… And this will prevent any trimming or slipping within the edit. So that concludes Scene Detection in Flame. Be sure to check out the other features, workflows… and updates to the Flame 2018 products. Comments, feedback and suggestions are always welcome and appreciated. Thank you for watching and please subscribe to
the Flame Learning Channel for future videos.

2 thoughts on “Scene Detection – Flame 2018

  1. In compare mode on viewing the timeline, please add some sort of auto scene detect/auto marker for the "difference" setting. That way if a difference is found it could auto mark rather than manually watching lots of black frames in the hope to find something different.

Leave a Reply

Your email address will not be published. Required fields are marked *