Video Frame Interpolator
Compare small blocks in adjacent frames to estimate motion, then project pixels from both images toward the midpoint. The tool encodes the synthesized frames into a real video file and avoids inventing motion across large scene changes.
Key features
- Double a 4, 6, 8 or 12 FPS sampling cadence for videos up to eight seconds
- Find block motion in both directions and synthesize midpoint frames
- Hold the earlier frame across a strong scene cut before showing the next real frame
- Play source and result side by side and inspect frame counts
- Export a playable MP4 or WebM according to browser encoder support
- Process locally and clean up on cancellation or file replacement
How to use
- Choose a 0.5–8-second MP4/WebM or generate the moving example.
- Select a base FPS close to the low frame rate of the source.
- Run interpolation to estimate midpoints and encode the new video.
- Play source and result to inspect scene boundaries and motion quality.
- Download the MP4 or WebM result. The exported video has no audio.
Use cases
- Experiment with a low-FPS product rotation clip
- Compare in-between motion in an animation draft
- Process a short demo that must stay local in the browser
- Inspect interpolation limits around scene cuts
Frequently asked questions
Does it merely duplicate frames?
No. It searches small blocks in adjacent images, estimates their displacement and projects pixels from both frames into an intermediate image. At a strong cut it deliberately holds the earlier scene rather than synthesizing a false transition.
Does it download an AI model or upload my video?
No. The algorithm is site-owned block-motion estimation with no learned model or weights. Video data stays in this browser and is not sent to a server or remote API.
Will every clip look smooth and accurate?
No. Fast motion, occlusion, rotation, complex texture, long travel and cuts can produce artifacts or ghosting. The block search reaches six pixels in each direction and the output long edge is at most 320 pixels.
Does it retain audio and the original resolution?
No. The current result is silent and downsized to a maximum 320-pixel long edge. Source audio, captions, metadata and original FPS are not preserved.
Why are clips limited to eight seconds?
Decoding, bidirectional block search and video encoding all run in the browser. The file cap is 64 MiB, duration 0.5–8 seconds and output cap 32 MiB to bound memory and processing time.
Which base FPS should I choose?
Choose a value near the source's low frame rate. For example, 12 FPS base creates a 24 FPS result. Sampling an already high-FPS clip at a lower rate may discard original motion information.
Privacy
Input and output stay in browser memory. There is no upload, remote model call, analytics event or browser storage. Source audio is not copied into the result.
Comments & questions