Skip to main content
Log inSign up

topaz/video-ai/frame-interpolation

Generate intermediate frames with Apollo, Chronos or Aion for smoother playback and up to 8× slow motion.

From $0.002/s
Run model

Workbench

Input

Frame interpolation
1/1

Drop a video here, or browse · 1 file · Max 90 MB · MP4

  • 1
1 / 1 files
Input must be a linked or uploaded MP4 of 1–60 seconds; duration violations are rejected before acceptance. At normal speed audio is kept, while tested slow-motion outputs have no audio. at least 1 characters
Style presetaffects price
ApolloApollo FastChronosChronos FastAion
- `apollo` — precise slow motion and rate conversion, default - `apollo_fast` — quick simple movement - `chronos` — natural real-world motion - `chronos_fast` — quick motion processing - `aion` — complex high-resolution extreme slow motion Default model: `apollo`; model selection affects the interpolation rate, with Aion the most expensive.
affects price
Available output tiers are `720p`, `1080p` (default), `2k` and `4k`; source framing is retained, and tier combines with model, fps and slowing to set cost.
Frame rateaffects price
30485060120
Output rate: `30`, `48`, `50`, `60` (default) or `120`; higher rates create more frames and increase cost. The source is read at 30 fps.
Slow Motionaffects price
1248
`1` (default) keeps normal duration; `2`, `4` and `8` extend it proportionally and raise the rate. Slowed test outputs omit audio; billable duration remains the source length.
Safety checkerOff

SpicyAPI does not run a safety checker on this model. Your prompt and media go to the model as written, and the model itself decides what comes back.

Checking your account before enabling submission and pricing.

Output

Ready for a new run
Example result
Extend
Video editing
Video upscaling

Related models and next tasks

Missing the model you need?

TOPAZ VIDEO AI API

Enlarge the picture, clean the noise, or slow the movement.

Key numbers

Enlargement models
21
Match faces, animation, interlacing or compression to a dedicated preset.
Repair choices
8
Select one of four Nyx treatments, deblur or three Hyperion HDR conversions.
Slower movement
×8
Extend action with generated frames while choosing a separate delivery frame rate.

Before you spend

Upscaling

Enlargement can leave compressed patches visible; a higher tier does not certify recovered source detail, and the two fixed-2× presets need a dimension check.

Video editing

Denoise, Deblur & HDR processes the clip at the requested resolution. Start with `nyx` for natural-looking noise treatment, or choose `deblur` for softness. Hyperion outputs HEVC (H.265), 10-bit HDR; browser previews may fail, so review the download in an HDR-capable player or editor. Other presets output H.264. The tested Hyperion 2.5 file is tagged PQ (`smpte2084`), while the Hyperion 1 sample is tagged HLG (`arib-std-b67`); check the returned file’s HDR metadata in your player or editor.

Frame interpolation

Input is read at 30 fps for frame interpolation. With `slow_motion` above `1`, duration increases and the tested output has no audio; ordinary enhancement and frame conversion without slowing retain the soundtrack.

What it does

Cyclists enlarged to 1080p by Proteus beside a matching crop of the compressed street recording.

Purpose-built models

Proteus

Matched cyclist crops: source left, Proteus 1080p right, with cleaner contours and some blocky patches remaining.

Inspect the result

Proteus · 1080p

A seated podcast speaker and microphone in a side-by-side source and Proteus enlargement.

Podcast frame: source left and Proteus 1080p right, showing the speaker and microphone at the same instant.

Two people among wine barrels, with noisy input and a 1080p Nyx crop showing reduced grain.

Purpose-built models

Nyx · 1080p

Matched cellar crops: artificially noised source left, Nyx 1080p right. Grain is much lower, hair clumps remain visible and background barrel texture is softened.

Inspect the result

Deblur · 1080p

A dancer shown before and after 1080p deblur processing, with softness still visible in the face.

Blurred dancer on the left and deblur 1080p on the right at the same time; the face remains soft.

Where it fits

  • Put the clip to work

    Prepare an old phone recording or an interview excerpt for a larger edit. Pick an enlargement preset suited to the source, then check skin, hair and compression marks before using the result in the final timeline.

  • Inspect the result

    Work on a dim interior, a blurred dance shot or a movement study with the matching tool. Keep noise treatment, HDR conversion and slow motion as deliberate choices so you can assess each result against its source.

Model reference

Topaz Video AI API reference

How to call the Topaz Video AI API

  1. Prepare a short take

    Identify whether the shot needs more pixels, cleaner texture or a change in playback speed.

  2. Choose the treatment

    Select the relevant endpoint before choosing a model; the three tools solve different problems.

  3. Review before delivery

    Download the processed take and check moving edges, facial detail and audio against your delivery requirements.

The short version

Trim to a useful passage of 1–60 seconds and send its MP4 through `video_url`. Start with `1080p` and the default preset for the chosen tool, keeping an untouched copy for review. A higher delivery tier cannot guarantee recovery of detail missing from the recording. Watch the full clip after processing, especially hands in motion and edges against busy backgrounds. For HDR, download the file to a compatible player or editor. For slow motion, plan a separate sound edit because the tested slowed outputs contain no audio track.

Topaz Video AI API questions

All Topaz models6

Video models3

Image models3