DaVinci Resolve 20: automate a pass, not an edit.

More tools, one timeline

Blackmagic Design announced DaVinci Resolve 20 at NAB in Las Vegas on 4 April 2025 and made the public beta available that day. Its release described more than 100 new features across the application, including AI IntelliScript, Animated Subtitles, Multicam SmartSwitch and Audio Assistant, alongside a keyframe editor, voiceover palettes, multi-layer compositing and updates to Magic Mask and Depth Map. That breadth is the important historical fact; it does not mean that every new control belongs in every edit.

Read the list as a map of handoffs. A text-derived timeline is an assembly proposal. Speaker-based multicam switching is a selection proposal. An audio assistant is a mix proposal. None is a story decision. The useful editorial question is not whether an automated pass is impressive; it is whether it gives the editor a legible starting point that can be checked, changed and abandoned without losing the project’s intent.

Keep the decisions visible

Start with a short written brief before asking a tool to make a first pass: the scene’s purpose, the material that may be omitted, the line that must land, and the point of view that should lead. For an interview sequence, a transcript can help locate themes, but it cannot decide whether a hesitation carries more meaning than a fluent answer. For a panel, speaker detection may locate who is talking, but it cannot know whether the strongest reaction is elsewhere in the frame.

Make an automation lane in the timeline rather than treating the returned result as the edit. Label the imported assembly or switch pass, preserve the original sync sequence, and keep a marker at every human intervention. Those markers turn revision notes into something a second editor can understand: “hold the reaction,” “do not subtitle this off-camera interruption,” “keep room tone through the cut.” They also reveal whether the assist reduced repetitive work or merely moved the same work into cleanup.

The release’s keyframe and voiceover additions point to a related discipline. A generated or assisted cut still needs intentional changes in emphasis, pace and sound. Put a temporary narration or guide track where it clarifies structure, then verify that picture changes have not made its timing misleading. Use keyframes to solve a specific transition or composite problem, not to hide a sequence whose internal logic has not been settled.

A review order that resists false finish

Review the pass with sound off first. Check screen direction, eyelines, continuity of gesture and whether each cut changes information, energy or point of view. Then listen without watching: are dialogue joins clean, does ambience bridge a deliberate cut, and has music disguised an awkward transition? Only after those two passes should captions, masks and graphics be judged in motion at normal speed.

This order matters because a polished automated result can look finished before it is coherent. Captions can be accurate yet obscure a face or arrive after a comedic beat; a clean mask can still drift where the eye is most likely to look. Use the output as evidence to make a decision, not as evidence that the decision has already been made. Blackmagic’s announcement is a product account, not a measurement of editorial quality, cost, rights clearance or labor saved. The craft remains in the accountable choices surrounding the tools.