The Vertical Video Demand Problem
Every client I work with in 2026 wants vertical video. Every single one. Even the ones who initially said they only needed a YouTube video come back a week later asking for TikTok and Reels versions. This is not a trend anymore. It is the default expectation.
The problem is that most professional video is still shot in 16:9 horizontal format. Cameras, monitors, and the entire filmmaking ecosystem is built around widescreen. When a client asks you to convert a carefully composed horizontal video into a vertical one, you are essentially asking to throw away 60 percent of the frame.
Without AI tools, vertical conversion means one of two options. Option one: crop to center and hope for the best, which results in awkward framing where speakers are cut in half and important elements are off-screen. Option two: manually keyframe the crop position for every shot to follow the action, which takes hours per video.
Auto reframe is the AI-powered third option. It analyzes each frame, identifies what is visually important (faces, motion, text, objects), and dynamically positions the vertical crop to keep those elements in frame. It is not perfect, but it gets you 80 to 90 percent of the way there with zero manual effort.
How AI Auto Reframe Works
Auto reframe uses computer vision to make cropping decisions frame by frame. Here is what happens under the hood:
Face detection and tracking. The AI identifies faces in each frame and prioritizes keeping them visible and centered. For talking-head content, this alone handles most of the reframing.
Motion tracking. When there are no faces (b-roll, product shots, landscapes), the AI tracks the dominant motion in the frame and positions the crop to follow it. A car driving left to right will keep the car centered in the vertical frame.
Saliency detection. Beyond faces and motion, the AI looks for visually salient areas: high contrast, bright colors, text, or unusual patterns. These get priority when no other tracking target is available.
Smooth transitions. The AI avoids jerky crop movements by smoothing the reframe path. Instead of jumping to a new position every frame, it uses gradual pans that feel like intentional camera movements.
I used to dread the "can we also get vertical versions" request. Now I budget 20 minutes per video for auto reframe plus manual cleanup, and I actually look forward to it. The AI handles the mechanical work, and I just need to fix the five to ten percent of shots where it makes a questionable creative choice. For a three-minute video, that is usually four or five adjustments.
Using Premiere Pro's Built-In Auto Reframe
Premiere Pro has included Auto Reframe since version 2020, and Adobe has steadily improved it. Here is how to use it effectively.
To apply Auto Reframe to an entire sequence: go to the Sequence menu, select Auto Reframe Sequence, choose your target aspect ratio (9:16 for vertical), and select a motion preset. Premiere creates a new sequence with the reframed content.
The motion presets are important:
- Default: Balanced tracking, works for most content
- Slower Motion: Smoother panning, better for interviews and static shots
- Faster Motion: More responsive tracking, better for action and sports
For most freelance work (interviews, brand videos, social content), the Default or Slower Motion preset produces the best results. Faster Motion tends to create distracting pan movements on static shots.
You can also apply Auto Reframe as an effect on individual clips rather than the entire sequence. This gives you more control because you can adjust the reframe settings per clip. Select a clip, go to Effect Controls, and modify the Auto Reframe tracking and position.
The main limitation of Premiere's Auto Reframe is that it operates purely on visual analysis. It does not understand the content or context of the shot. It cannot tell the difference between a speaker who is important and a background extra, or between a product that should be centered and a decoration that can be cropped out.
Semantic Reframe: The Next Level
Semantic reframe goes beyond face and motion tracking. It understands what is happening in the scene and makes cropping decisions based on meaning, not just visual saliency.
For example, in a cooking video, semantic reframe understands that when the chef is chopping vegetables, the important element is the hands and the cutting board, not the chef's face. A basic auto reframe would track the face and crop out the cooking action. A semantic reframe tracks the relevant subject.
Wideframe's semantic analysis enables this kind of intelligent reframing. Because it understands the content of each shot, it can make smarter cropping decisions that preserve the meaning and visual storytelling of the original horizontal frame.
Other examples where semantic reframe outperforms basic auto reframe:
- Two-person conversations: Instead of bouncing between two faces, semantic reframe holds a wider crop that includes both speakers during dialogue
- Product demonstrations: Tracks the product and the demonstrator's hands, not just the person's face
- Text on screen: Recognizes and prioritizes on-screen text (titles, graphics, slides) that would be cropped out by face-only tracking
- Wide establishing shots: Understands that some shots are meant to show scale and applies a slow pan across the scene rather than zooming into one element
Handling Tricky Shots That Break Auto Reframe
Auto reframe handles most shots well, but some compositions are inherently incompatible with vertical cropping. Here is how to handle the common problem cases.
Side-by-side comparisons. When two elements are placed next to each other horizontally, vertical reframe has to choose one or the other. Solution: do not auto reframe these shots. Instead, manually create a stacked vertical layout with one element on top and the other on the bottom.
Wide group shots. A group of six people standing in a horizontal line cannot fit in a vertical frame no matter how smart the AI is. Solution: use a slow pan across the group rather than trying to fit everyone in frame, or replace with individual shots if available.
Screen recordings. Horizontal computer or phone screens in a horizontal video look tiny when reframed vertically. Solution: zoom into the relevant portion of the screen and manually position it. For screen recording content, consider recording in vertical format from the start.
Text-heavy slides. Presentation slides with small text become illegible in vertical format. Solution: split the slide into multiple vertical frames that show sections of the text at readable size, or recreate the key information as vertical-native graphics.
Panning shots. Horizontal pans look strange in vertical because the motion path fights with the reframe's own tracking. Solution: lock the reframe position during pans and let the original camera movement provide the motion.
Step-by-Step Vertical Conversion Workflow
Maintaining Quality After Reframing
Reframing inevitably involves cropping, and cropping means you are using a smaller portion of the original frame. Here is how to maintain visual quality through the process.
Start with the highest resolution source. If you shot in 4K (3840x2160), your vertical crop uses about 1080x1920 pixels of the frame, which is exactly 1080p. You lose nothing. If you shot in 1080p, the same vertical crop uses about 607x1080 pixels, which is noticeably soft. This is why shooting in 4K matters even if your final delivery is 1080p. It gives you room to crop without quality loss.
Avoid digital zoom on top of reframe. If the auto reframe already crops the frame, adding further digital zoom pushes beyond the source resolution and creates visible softness. If you need to zoom in further, consider AI upscaling first.
Watch for compression artifacts in dark areas. The smaller frame of vertical video can make compression artifacts more visible, especially in dark or low-contrast areas. Export at a higher bitrate than you would for horizontal video.
Check safe zones. TikTok, Instagram, and YouTube Shorts overlay UI elements (username, caption text, action buttons) over the bottom and top of the video. Keep important visual content in the center 80 percent of the frame vertically. Platform-specific safe zone guides are available online and worth following.
Platform-Specific Reframe Considerations
Each vertical platform has slightly different conventions that affect how you should reframe your content.
TikTok: Fast-paced content works well. Use the Faster Motion preset for dynamic content. TikTok viewers expect visual energy, so do not be afraid of more aggressive reframe movements. Keep the most important content in the top two-thirds of the frame because captions and UI cover the bottom.
Instagram Reels: Slightly more polished than TikTok. The Slower Motion preset often works better here. Instagram users expect higher production value. Ensure your auto-reframed content still looks intentionally composed, not accidentally cropped.
YouTube Shorts: YouTube's player has the least UI overlay, giving you more usable screen space. Shorts can be up to 60 seconds and benefit from compositions that use the full vertical frame. Consider adding vertical-native graphics rather than reframing horizontal ones.
LinkedIn: Supports both 9:16 and 1:1 vertical formats. For professional content, 1:1 square is often a better choice than full vertical because it takes up more feed space on desktop while still looking good on mobile. Auto reframe to 1:1 loses less of the original frame and requires fewer corrections.
My advice for anyone starting with auto reframe: do your first three or four projects with careful manual review of every shot. You will quickly learn which types of shots reframe well and which ones always need manual fixes. After that initial learning period, you can review more efficiently because you already know where the problems will be. Within a month of using auto reframe regularly, I could predict with 90 percent accuracy which shots would need manual adjustment just by looking at the horizontal version.
Stop scrubbing. Start creating.
Wideframe gives your team an AI agent that searches, organizes, and assembles Premiere Pro sequences from your footage. 7-day free trial.
Frequently asked questions
In Premiere Pro, go to Sequence > Auto Reframe Sequence, select 9:16 as the target aspect ratio, and choose a motion preset. The AI generates a new vertical sequence by tracking faces and motion. Review the result and manually fix any shots where the framing is incorrect.
Auto reframe crops the frame, using a smaller portion of the original resolution. If you shot in 4K, the vertical crop produces a full 1080p frame with no quality loss. If you shot in 1080p, the crop results in about 607x1080 pixels, which will appear softer. Shooting in 4K gives you reframe flexibility without quality loss.
For interviews and talking-head content, use the Slower Motion or Default preset. These produce smooth, subtle reframe movements that keep the speaker centered without distracting camera-like panning. The Faster Motion preset is better for action and sports content.
Yes. In Premiere Pro, you can apply Auto Reframe as an effect on individual clips through the Effect Controls panel. This gives you per-clip control over tracking settings and crop position, which is useful when different shots need different reframe approaches.
Semantic reframe uses AI that understands the content and meaning of each shot, not just faces and motion. It makes smarter cropping decisions, like tracking a product being demonstrated rather than just the demonstrator's face. Tools like Wideframe offer semantic reframe for more intelligent vertical conversions.