A ceramic artist films one continuous vertical clip of a spinning bowl on a pottery wheel. The workshop later needs the same scene in a wide training-video frame. Stretching the portrait image would distort the bowl; cropping it to landscape would remove the hands. Runway Expand Video in Edit Studio instead synthesizes space around the original frame. That extra area is not captured evidence, however, and a moving hand near the seam can produce an unconvincing extension. This workflow outlines a controlled source, one keyframe review, a cautious prompt, and a frame-by-frame release check. The pottery illustrations are original teaching diagrams, not output produced in Runway.
Check the source before changing its aspect ratio
Use one continuous, rights-cleared shot rather than a montage. Runway’s Expand Video documentation says inputs should use conventional aspect ratios, be longer than two seconds and shorter than thirty seconds, fall between 480p and 1080p, and run at twenty-four to thirty frames per second. Videos above thirty seconds may be trimmed, and high frame rates may be downsampled. The tutorial recommends a shot without cuts because an extended background has to remain plausible as the camera and subject move. For the pottery example, note exact wheel direction, hands, bowl size, and the edge of the table. Archive the source untouched and log how much of it you intend to reuse.
Set a meaningful landscape composition instead of a generic fill
Open Edit Studio in Runway Apps, upload the permitted source, and choose a wider aspect ratio. In the example a narrow center portrait stays intact while new scenery is proposed on left and right. Sketch the desired crop before running AI: a clean worktable edge on the left, softly lit shelves on the right, and ample empty room for a lesson title. Avoid asking for extra tools hanging over the hands, extra pottery pieces, or an assistant who did not appear in the original shoot. Keep the original bowl and hands central. A brief prompt can guide what appears in the new region, but leaving it empty permits automatic context-based fill.
Review the keyframe before paying for motion generation
Runway describes an image-based keyframe step: submit an edit, then inspect prompt versions, reuse a prompt, or open the Preview before-and-after comparison. Do not infer success from the small generated thumbnail. Enlarge the preview and compare the original center with the output. Check whether the artisan’s fingers are still attached to the same hand, whether wheel geometry remains circular, and whether shelves are shaped plausibly. A newly imagined wall clock or written sign may look authoritative without reflecting the real location. Reject any version that changes the original pottery process or invents important instructional details. Label candidate images as speculative compositions, not real photographs of the workshop.
Choose a single generated edge design and save its prompt
Create a small version register with original file identifier, target ratio, candidate number, prompt, reviewer, and notes. If the right shelf looks stable but the left tabletop sprouts unexplained objects, change just the left-edge wording. Comparing several revisions is useful only when a person identifies the error each revision is intended to fix. Keep a screenshot or copy of the source framing; this is the visual anchor during every approval pass. The diagram shows a conceptual three-column edge study rather than actual Runway generations. Avoid choosing a version because its wide scenery is spectacular when the resulting action no longer looks like the source lesson.
Generate video after the keyframe composition passes review
Once a keyframe is acceptable, choose Generate Video. Runway states that Aleph 2.0 uses the selected keyframe to guide the motion extension; no separate motion prompt is required for this step. Review the newly synthesized area through the entire clip rather than the first second only. Vertical edges that look fine when stationary may drift when the subject passes across them. Check the back edge of the table, the continuity of the shelf shadows, and reflections on glazed clay. Inspect the first, middle and last frames, followed by the entire sequence at speed. The technical workflow enables generation; it does not guarantee identity or physical continuity.
Watch the original-versus-generated border during movement
Put the original portrait reference in a center-frame overlay for editorial checking. Follow the left and right borders while the wheel spins, hands move, and the potter leans. A common risk is temporal shimmer: a shelf handle appears in one frame and dissolves later, or the table seems to breathe. Examine whether newly generated objects cast shadows that track the workshop light. If the extended space contains words, labels, people or safety signs, verify them against source documentation or remove them. Do not publish a synthetic wall notice as if it was visible in the original room. The acceptance check is continuity, not only aesthetic plausibility.
Plan titles, graphics and accessibility around the safe area
The widened frame can create space for captions and a lesson title, but the generated scenery should remain subordinate to the actual hands and bowl. Place titles in a consistent region and inspect both desktop and mobile viewing. A widescreen export may eventually be re-cropped for vertical feeds; avoid putting essential text at extreme edges that disappear in the next format. Keep subtitles on a clearly contrasted backing rather than directly over busy shelves. Record which pixels are original and which were synthesized in the production notes. If viewers are learning a safety-sensitive technique, any tool placements or steps must be verified from the source footage, not from outpainted decorations.
Approve an honest export and keep a rollback version
Compare a full-size render with the camera original on a calibrated or at least consistent display. Confirm unchanged center content, plausible edge motion, correct playback speed, audio continuity, caption legibility, and the intended aspect ratio. Do not confuse a successful keyframe image with a successful moving clip. Retain the source, prompts, selected version, and final editorial approval so later reviewers can identify the generated region. For documentary work, disclose material synthetic scene extension where it could affect interpretation. For a straightforward training graphic, a concise disclosure in production notes may be enough. The original source must remain available for any future correction.
How We Research
We reviewed Runway’s October 2026 Edit Studio Expand Video instructions and the Runway Gen-4 help page documenting the handoff into Edit Studio on October 28, 2026. We did not upload pottery footage or run Runway generation. All aspect-ratio crops, shelf extensions and keyframe examples are conceptual original artwork. Feature availability, model credits and output quality may vary.
Frequently Asked Questions
The documented goal is to synthesize new space outside the frame rather than crop the original footage, but output still needs visual QA.
Runway recommends a cut-free, single-shot input; a multi-shot sequence may be inconsistent and should be split first.
The documented Edit Studio tutorial specifies longer than two seconds and shorter than thirty seconds, with long files trimmed.
Runway says Aleph 2.0 uses the selected keyframe to guide the generated video at that stage.
Not without appropriate disclosure and verification; synthetic shelves or signs were never captured by the original camera.