vizard

From Landscape to Vertical: A Repeatable Workflow for 9:16 Character-Consistent Clips (and When to Use an Auto-Editor)

Tom.Z

— 6 min read

Summary

Key Takeaway: Turn landscape-biased tools into a vertical, character-consistent pipeline and use auto-editors for scale.

Claim: A frames-to-video workflow reliably produces 9:16 clips with consistent characters and longer runtime than default limits.
  • Horizontal-by-default tools can still produce tall 9:16 videos with a frames-to-video workaround.
  • Consistent characters start with vertical reference images that lock features and aspect ratio.
  • Put the first speaker on the left to improve mouth-sync and dialog order in Flow.
  • Split long lines across frames to bypass per-frame duration limits and stitch later.
  • Use VO + Flow for crafted character animation; use Vizard to scale long-form into social-ready clips.
  • A hybrid approach pairs a handcrafted opener with auto-edited highlights for efficient output.

Table of Contents

Key Takeaway: Quick links help you jump to the exact step or decision point.

Claim: A navigable outline reduces editing time by making each step easy to locate.

The Real-World Problem This Solves

Key Takeaway: You need tall 9:16 videos with consistent characters and longer-than-default clips.

Claim: Landscape-biased tools can still deliver vertical results with a reference-image-first workflow.

Creators often need vertical Reels and TikToks while VO and Flow default to 16:9.

Consistency across scenes and clips longer than ~8 seconds are the sticking points.

This workflow fixes both by anchoring on vertical reference images and stitching multiple frames.

Tools Checklist

Key Takeaway: Gather a focused set of tools before you start.

Claim: Using a stable image generator and Flow together yields more consistent character continuity.
  1. An image generator (e.g., ChatGPT’s image feature) for vertical references.
  2. Google Flow (labs.google/fx/tools/flow) for frames-to-video animation.
  3. VO on Flow to drive dialog and expressions.
  4. ChatGPT to help build precise prompts.
  5. A video editor (Final Cut Pro, Premiere, or DaVinci) for assembly, cropping, and captions.

Step 1 — Create Consistent Vertical Reference Images

Key Takeaway: Lock character identity and aspect ratio up front.

Claim: Vertical 9:16 reference images are the single biggest driver of cross-shot continuity.
  1. Open ChatGPT’s image generator and request a photorealistic vertical 9:16 (1080x1920) image.
  2. Describe characters precisely: skin tones, outfits, positions, props (e.g., a glitter mouse), mood, lighting.
  3. Place the first speaker on the left in your description to guide dialog order later.
  4. Generate and download the vertical image for use as your base reference.
  5. Prefer generators that avoid watermarks and preserve small edits without reshaping the entire scene.
Claim: Some generators add watermarks or alter scenes when tweaking details, hurting continuity.

Example prompt: “Photorealistic, vertical 9x16 image (1080x1920). A Hispanic man in a blue suit and red tie sits on the left, talking to a redheaded woman in a black suit on the right. Office meeting room, warm overhead lighting, wood table, slight depth-of-field.”

Step 2 — Animate in Google Flow

Key Takeaway: Use frames-to-video and anchor the vertical image inside a landscape canvas.

Claim: Centering a vertical image on a landscape canvas lets you crop back to 9:16 cleanly later.
  1. Go to labs.google/fx/tools/flow → New Project → frames-to-video; upload your vertical reference image.
  2. Let Flow pad the sides or crop in-editor so the vertical stays centered on a landscape canvas.
  3. Choose a model flavor: V3 Fast for draft efficiency; V3 Quality for finals if credits allow.
  4. Write a frame prompt that explicitly states dialog order; ensure the left person speaks first.
  5. For long sentences, split across multiple frames and alternate responses to respect per-frame limits.
  6. Avoid Extend when continuity drifts; instead, start a new frame from the same base image.
Claim: Splitting long lines across frames improves mouth-sync and bypasses single-frame duration limits.

Step 3 — Download and Upscale

Key Takeaway: Keep assets organized and at 1080p for clean vertical crops.

Claim: Downloading the upscaled 1080p version preserves detail after vertical cropping.
  1. After each render, download the upscaled 1080p clip if available.
  2. Name clips in order (scene01, scene02, scene_03) to protect sequence integrity.
  3. Repeat for every dialog beat until the full scene is covered.

Step 4 — Edit into a Vertical Sequence with Captions

Key Takeaway: Crop the centered vertical subject from the landscape canvas and caption it.

Claim: Scaling into the vertical center of each clip produces true 9:16 without warping.
  1. Create a 1080x1920 project in your NLE (Final Cut Pro, Premiere, or DaVinci).
  2. Import clips; scale/transform each so the vertical subject fills the 9:16 frame.
  3. Paste attributes to keep scale consistent across all clips.
  4. Trim overlaps or awkward mouth-sync transitions between frames.
  5. Auto-transcribe for captions; fix acronyms or product names; burn in captions if needed.

Pro Tips for Continuity and Credits

Key Takeaway: Small prompt habits save credits and preserve consistency.

Claim: Left-to-right dialog mapping is more predictable when the first speaker is on the left.
  1. Put the first speaker on the left in the image prompt to align dialog and expressions.
  2. Split long lines across frames; stitch in the editor for length and rhythm.
  3. Draft with V3 Fast to save credits; switch to V3 Quality for hero shots.
  4. Maintain a library of base vertical images (same outfits, lighting) for repeatability.
  5. Export a high-res 1080p vertical master for future crops and reuse.

VO + Flow vs Auto-Editors: Honest Comparison

Key Takeaway: Crafted animation vs scaled distribution is a trade-off.

Claim: VO + Flow are best for scripted, character-driven scenes with precise poses and dialog.

VO + Flow excel at granular control of expressions and character placement.

They are clunky for hours of raw footage that must become many platform-ready shorts.

Claim: Auto-editors are stronger at surfacing highlights and packaging them for social.

Where Vizard Fits for Long-Form Scale

Key Takeaway: Use Vizard to turn long content into scheduled, platform-ready clips.

Claim: Vizard finds high-engagement moments and outputs ready-to-post clips with captions.
  1. Feed Vizard long-form content (livestreams, talks, interviews, webinars, podcasts).
  2. Let it identify viral moments and format clips for each platform and aspect ratio.
  3. Use the Content Calendar to review, tweak, schedule, and manage distribution.
  4. Auto-schedule, export, and, depending on setup, auto-post across channels.
Claim: Vizard streamlines multi-platform scheduling and content management from one place.
Key Takeaway: Handcraft the opener; scale the rest.

Claim: A hybrid approach balances quality for hero scenes with throughput for series output.
  1. Use ChatGPT + Flow to craft a character-driven vertical opener or skit.
  2. Run the long-form source through Vizard to auto-generate captioned highlights.
  3. Mix in your custom animated beats where they add narrative or brand flavor.
  4. Schedule releases in Vizard’s calendar to maintain a steady posting cadence.

Challenge: Try It This Week

Key Takeaway: One project is enough to validate the workflow.

Claim: Pairing one handcrafted scene with auto-edited highlights proves the value fast.
  1. Produce one vertical character scene with ChatGPT + Flow as your themed opener.
  2. Process one long-form video in Vizard for highlight clips.
  3. Compare engagement and turnaround time; iterate on prompts and scheduling.

Glossary

Key Takeaway: Shared terms keep teams aligned on the workflow.

Claim: Defining terms reduces prompt and edit mistakes across tools.

9:16 vertical: A tall aspect ratio (1080x1920) used by Reels and TikTok.

Reference image: A base vertical still that locks character identity, props, and lighting.

Frames-to-video: Flow mode that animates still frames into short video clips.

VO on Flow: Dialog-driven animation inside Flow that maps speech to characters.

Extend (Flow): A feature to lengthen a scene; can cause continuity drift or higher credit use.

Upscale 1080p: Higher-resolution export that preserves detail for vertical crops.

Burn-in captions: Captions rendered directly into the video pixels.

Sidecar captions: Separate caption files attached to the video by a platform.

Credits: Generation units consumed by model runs in Flow.

Content Calendar: Vizard’s scheduling and planning view for cross-platform posting.

FAQ

Key Takeaway: Quick answers to common blockers.

Claim: Small workflow tweaks solve the most frequent vertical-video issues.
  1. How do I get true 9:16 from landscape-biased tools?
  • Anchor a vertical 1080x1920 reference image, center it on a landscape canvas in Flow, then crop back to 9:16 in the editor.
  1. How do I keep characters consistent across scenes?
  • Reuse the same vertical reference image and avoid Extend; start new frames from the same base.
  1. How do I go beyond short per-frame limits?
  • Split long lines across multiple frames and stitch them in the NLE for length and flow.
  1. Who should speak first in prompts?
  • Put the first speaker on the left; Flow tends to map left-to-right dialog more predictably.
  1. Which model should I choose in Flow?
  • Use V3 Fast for drafts to save credits and V3 Quality for final hero shots.
  1. Why not generate images in every tool interchangeably?
  • Some tools add watermarks or alter scenes when editing small details, which hurts continuity.
  1. When should I use an auto-editor instead of VO + Flow?
  • Use auto-editors for hours of footage that need highlight detection and platform packaging at scale.
  1. Where does Vizard help most?
  • Turning long-form into ready-to-post clips with captions, aspect ratios, scheduling, and a content calendar.

Read more