Depth-Aware Video Frame Interpolation (CVPR 2019), the classic research code.
Legacy: 2019 method; RIFE/FILM-based tools are more practical now.
Worth knowing
Price: Open source (MIT)
License: MIT
Status: sourced. We have not run this ourselves yet (no "Tested" badge).
Quality bar (computed by code): 4/5 (primary source ✓, reproducible ✗, new ✓, example output ✓, safe ✓)
Get Atlas Dispatch
The best new AI video workflows and prompts, weekly.
Weekly: 5 new prompts and workflows. Unsubscribe anytime.
From the source
Setup. Research code for Depth-Aware Video Frame Interpolation (CVPR 2019). It was tested on Ubuntu 16.04 with Python 3.6, CUDA 9.0 and PyTorch 1.0.0, and you compile custom CUDA extensions with gcc 4.9 and nvcc 9.0. That's an old stack, so expect setup work on current GPUs. The README links a Google Colab demo. MIT license.
Inputs. Video frames. For slow motion, set time_step (0.25 = 4x, 0.125 = 8x, 0.1 = 10x; the authors mention 0.01 for 100x 'for fun').
What the source says about results. The authors report state-of-the-art results on the Middlebury benchmark at publication. The method uses depth cues to handle occlusion, sampling closer objects preferentially when synthesizing in-between frames.
Summarised from github.com/baowenbo/DAIN, read 2026-10-11. Claims are the source's, not our tests.
For your agent
Give your AI agent a video director
Connect Claude, ChatGPT, Cursor, Gemini or Grok Bot to Atlas, and your agent can plan shots, pick models and price them from the data on this site. Free. No sign-up or API key.
▶ Play demoYouTube · loads on click