NEW AI Film Tools: The Biggest Updates for Artists
Using depth maps and pose skeleton inputs gives filmmakers significantly more physical accuracy and camera tracking control than standard image-to-video referencing.
What this lesson covers
Curious Refuge demonstrates how depth maps and pose skeleton rigs can be used as control inputs for video generation in Seedance 2.0. The tutorial compares raw reference video against depth map extraction tools and Segment Anything subject isolation.
Key takeaways from the creator
AI-extracted notes, not independently verified product claims. Timestamp links let you check each point in the original video.
- 00:30 ↗
Depth maps isolate spatial structure from reference video while stripping away unwanted background details prior to AI video generation.
- 03:52 ↗
fal.ai provides access to Depth Anything Video for automated, rotoscope-free depth map extraction.
- 04:40 ↗
Framethrower extracts both depth maps and OpenPose skeleton rigs from reference footage to supply additional pose structure.
- 05:31 ↗
Standard depth maps delivered better physics and physical weight consistency in Seedance 2.0 than skeleton pose overlay maps.
- 07:48 ↗
Meta's Segment Anything tool enables creators to isolate and color-code specific characters for precise multi-subject generation guidance.
Workflow outlined in the video
- Generate base character sheets and environment background images using AI image generators.
- Run performance reference video through fal.ai's Depth Anything model to automatically output a video depth map.
- Load the depth map video, character images, and prompt parameters into Seedance 2.0 using the Omni reference model.
- Utilize Meta's Segment Anything tool to isolate and color-code individual subjects when directing multi-character shots.
- Compare depth map outputs against direct video reference renders to select the optimal balance of physics and detail.
Before you use this workflow
These notes describe the source video at its publication date. Model access, pricing, connectors and interfaces may have changed. Check the original source and the provider’s current documentation before installing an add-on, connecting an account or spending credits.
ReelStack has not independently tested this workflow. Preview one representative shot and check motion, continuity and output quality before applying it to a full production. No result, cost saving or model capability is guaranteed.
Explore the tools
How to interpret the numbers
27,736 views captured 22 October 2026. ReelStack recommendation score: 50/100. Formula: radar-v5. This is ReelStack’s calculation, not a YouTube rating or a measure of factual accuracy.
Show the calculation
- 45% performance: 1.74× views vs 9 other channel videos observed at a similar age. View-evidence factor 97% (views / (views + 1,000))
- 25% freshness: 1/100 with a 14-day half-life
- 20% momentum: 12 views/day vs 5 channel baseline (3 comparable uploads), with the same view-evidence factor
- 10% engagement: 85/100; likes + 4× comments, smoothed with a 500-view neutral prior