Runway added an enhance_frame_rate option on September 17, 2026, with targets up to 120 FPS. That makes it easier to change FPS, but it does not make every converted clip more truthful or more useful. Interpolation creates frames between recorded moments; fast hands, crossings, text, and textured backgrounds can expose where those invented frames fail.
Prepare and review your FPS test in APOB’s AI Video Editor
Use one owned 12-second clip and build a matched 24/30/60/120 FPS ladder. Runway’s dated API changelog currently lists those rates among supported targets, a 300-second maximum input, and billing of one credit per two input seconds. Those are vendor statements for the Runway Dev surface, not guarantees about quality and not features claimed for APOB.
This is a change video FPS test, not a generation benchmark. The untouched source remains the reference at every stage, and the same acceptance sentence applies to each rung.
Record the video frame rate exactly as the file reports it; 29.97 and 30 are not interchangeable labels for an audit.
Freeze one motion plate and an acceptance rule
Choose a clip that is difficult enough to reveal interpolation errors but ordinary enough to resemble real delivery. A useful plate includes a person passing a product between hands while the camera pans past a shelf with small labels. Keep the untouched source and export its metadata before submitting anything.
Source cadence
Record container, codec, dimensions, duration, nominal frame rate, total frame count, variable- or constant-frame-rate status, and audio presence. Use a tool that reads file metadata rather than trusting a player label. If the source is 29.97 FPS, do not round it to 30 in the manifest.
Extract frame thumbnails around three events: hand overlap, product crossing a shelf edge, and the fastest camera movement. These markers let every reviewer inspect identical moments.
Play the source in at least two players and step through the marker region. If reported cadence differs, resolve variable-frame-rate interpretation before conversion. Otherwise a later mismatch may come from metadata rather than new frames.
Motion stressors
List what can break before conversion: thin fingers, partial occlusion, motion blur, repeating patterns, reflections, rapid pans, and objects entering the frame. Assign each stressor a timecode. Do not replace the plate after seeing an ugly result; the failure is the point of the test.
Protected details
Mark details that must not change: product silhouette, visible label letters, finger count, jewelry, shelf lines, facial geometry, and audio sync. A 120 FPS video that feels smooth but corrupts the product or hands fails.
Acceptance sentence
Write one decision rule before processing: Choose the lowest target rate that passes hand shape, edge integrity, text legibility, cadence, audio sync, and the intended delivery path. This prevents “more frames” from becoming an automatic winner.
Source field | Value |
|---|---|
File ID | ___ |
Duration | ___ |
Native FPS | ___ |
Frame count | ___ |
Dimensions / codec | ___ |
Stress timecodes | ___ |
Protected details | ___ |
Build the four-rung FPS ladder
Submit the same source to the same Runway endpoint and enhance_frame_rate model. Change only target rate. Record request ID, date, wait time, accepted/rejected status, duration, file metadata, and credit charge. Confirm current availability in the Runway model documentation before repeating the test.
24 FPS baseline
The 24 FPS output is not “bad” by definition. It is a delivery baseline for a cinematic cadence and shows whether conversion itself changes frames even when the target is relatively low. Compare it with the native source one frame at a time around the three stress markers.
30 FPS delivery
Use 30 FPS as a common delivery candidate, not a universal standard. Check whether the platform or editor expects 29.97 instead. An avoidable fractional-rate mismatch can create cadence artifacts unrelated to the interpolation model.
60 FPS motion
The 60 FPS video should make motion easier to inspect and can support smoother playback or a modest slow-down. Judge the inserted frames, not only the subjective smoothness. Count distortions at the predetermined timecodes.
Check full-speed playback first, then half speed. A defect visible only while stepping frames may be acceptable for one delivery, while a repeating wobble at normal speed is not.
120 FPS stress case
Treat 120 FPS as the stress case. It asks the system to synthesize many more intermediate moments. Test full-speed playback and a four-times slow-motion retime, then inspect the same source action. Do not assume a display, browser, social platform, or editing timeline will preserve all 120 frames.
Archive the 120 FPS video as a test output, not automatically as the master. Confirm the exported file truly reports the requested cadence and duration before judging it.
Target | Request | Wait | Credits | Output FPS | Frames | File size | Accepted |
|---|---|---|---|---|---|---|---|
24 | ___ | ___ | ___ | ___ | ___ | ___ | Yes / No |
30 | ___ | ___ | ___ | ___ | ___ | ___ | Yes / No |
60 | ___ | ___ | ___ | ___ | ___ | ___ | Yes / No |
120 | ___ | ___ | ___ | ___ | ___ | ___ | Yes / No |
For a 12-second input, the changelog’s listed rate implies six credits for each conversion, subject to the current API behavior and account. Verify the returned usage rather than treating arithmetic as a receipt.
Inspect interpolation failures frame by frame
Place the four outputs in synchronized tracks. Export contact sheets spanning the same half-second around each marker. Two reviewers should score without seeing the target rate label.
Keep the matched clips and annotations together in an APOB AI Video Editor review sequence before deciding which file advances to delivery.
Occlusion errors
Watch the handoff as fingers pass in front of the product. Flag doubled fingers, temporary holes, fused edges, object swaps, and a hand that changes depth order. Note the exact frame and whether the artifact appears for one frame or persists.
Edge tearing
Inspect the product against shelf lines. Interpolation can create a split edge when foreground and background move at different speeds. Count edge discontinuities and identify whether the source already contains blur. A converter should not be blamed for information the camera never captured.
Texture shimmer
Review hair, fabric, shelf patterns, and reflections at normal speed and frame-by-frame. Shimmer may be subtle in a local player but distracting after compression. Record the viewing device and playback rate.
Text and logo damage
Zoom into label letters and logos only if you have permission to use them. Reject frames that invent strokes, duplicate characters, or warp the mark. A horizontal product movement is a particularly useful check because small text crosses many backgrounds.
Timecode | Stressor | 24 | 30 | 60 | 120 | Reviewer agreement |
|---|---|---|---|---|---|---|
___ | Occlusion | ___ | ___ | ___ | ___ | ___ |
___ | Edge | ___ | ___ | ___ | ___ | ___ |
___ | Texture | ___ | ___ | ___ | ___ | ___ |
___ | Text/logo | ___ | ___ | ___ | ___ | ___ |
Score motion clarity and artifact burden separately. Otherwise the smoothest version can hide the largest number of invented defects.
Test each version in a real delivery path
Local playback is only one environment. Put each accepted file through the route it is meant to serve, keeping the account, upload settings, and caption state constant.
Social playback
Upload private or unlisted where possible. Record whether the platform recompresses, changes frame rate, limits resolution, or alters audio sync. View on a phone over normal network conditions. If 120 FPS becomes 60 or 30 after upload, judge the delivered file rather than the local master.
Slow-motion retime
Place 60 and 120 FPS versions in a 30 FPS timeline and slow them with frame blending disabled. Check whether the extra frames produce usable motion or merely reveal more artifacts. Compare against an optical-flow retime of the source only if you document it as a separate method.
Editorial handoff
Import the candidates into the editor used by the next person. Confirm duration, FPS interpretation, audio sync, color, naming, and proxy behavior. The APOB AI Video Editor can hold the practical review and assembly path, while high-end retiming or frame repair may still require a dedicated finishing tool.
File-size effect
Measure file size and bitrate; do not infer them from FPS alone. Encoding settings can dominate. Record upload time, storage, proxy generation, and export time if those costs matter to the team.
For creators already testing Runway generation, the APOB Runway AI Video Generation page is a relevant creation entry point. It does not mean APOB exposes Runway’s enhance_frame_rate API. Keep generation and conversion claims separate.
Choose convert, keep, or reshoot
Combine motion clarity, artifact burden, delivery survival, and operating cost. Then choose the lowest rate that passes the actual job.
Keep source FPS
Keep the source when it already meets delivery, interpolation adds defects, or the final platform discards extra frames. “No conversion” is a valid result and often the cleanest one.
Convert once
Convert once when a specific delivery or retime requirement justifies it and one target passes. Preserve the native source, processing receipt, and accepted output. Avoid converting the converted file again.
Use 120 FPS selectively
Use 120 FPS when the final workflow preserves it or needs the additional frames for an approved slow-motion treatment—and only when the protected details survive. It is a tool for a defined motion problem, not a general quality badge.
Reshoot trigger
Reshoot when source motion blur, rolling shutter, occlusion, missing detail, or a critical logo prevents a clean result. Interpolation cannot recover a moment the camera never recorded. A higher shutter speed or true high-frame-rate capture may be the honest fix.
Decision | Choose when | Reject when |
|---|---|---|
Keep source | Delivery passes; conversion adds no value | Required cadence or retime fails |
Convert to 30/60 | Specific delivery passes with low artifacts | Platform discards benefit or details warp |
Convert to 120 | Slow motion/high-rate master is required and verified | Stress frames invent visible errors |
Reshoot | Source information is missing | A verified conversion already solves the job |
Run the ladder on one owned clip, preserve the metadata, and publish the time-coded failures alongside any preferred output. Higher FPS can be useful. It is not automatically higher quality.
Sources
Be the first to like this.
No credit card needed
