I Tried Every AI Video Generator - Full Tierlist
Provides AI filmmakers with a direct head-to-head performance comparison of current leading video models using identical reference assets, exposing common failure modes like hallucinated props and physics glitches.
What this lesson covers
Dan Kieft tests six AI video models—Wan 3.0, MiniMax H3 Max, Flux 3, Gemini Omni 1.1 Flash, Kling 4.0 Flash, and Seedance 2.5—across eight standardized testing rounds. The creator benchmarks performance on prompt adherence, consistency in UGC ads, motion physics, and error handling to rank each model.
Key takeaways from the creator
AI-extracted notes, not independently verified product claims. Timestamp links let you check each point in the original video.
- 02:11 ↗
Standardizing benchmark criteria to fixed durations (10 seconds), 16:9 aspect ratio, and identical prompt references isolates model capabilities from prompt variations.
- 03:36 ↗
Wan 3.0 demonstrated strong multi-reference adherence, preserving character features, sword props, and window reflections across shot cuts.
- 06:07 ↗
Seedance 2.5 distinguished itself in complex environmental interactions, accurately solving real-world physics like door swings and handle placement that other models failed.
- 08:17 ↗
Using character reference sheets in Flux 3 can cause the model to render the literal reference image layout directly as a frame inside the generation.
- 09:13 ↗
Prompts requesting UGC smartphone aesthetics (e.g., 'filmed on iPhone') can prompt models like Gemini Omni and Flux to render an unwanted physical phone into the scene.
Workflow outlined in the video
- Benchmark new video generation models using consistent 10-second, 16:9 parameters with identical multi-image references to ensure fair comparison.
- Check prop and reflection persistence across continuous cuts when testing multi-shot narrative prompts.
- Remove literal device descriptors like 'filmed on iPhone' from camera prompts if a model hallucinates handheld phone borders into the video.
- Inspect subtle environmental physics, such as door opening directions and object interactions, to detect telltale AI generation flaws.
- Review character turnaround sheet inputs when using Flux models to ensure the reference frame does not flash on screen.
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.
How to interpret the numbers
38,917 views captured 30 October 2026. ReelStack recommendation score: 48/100. Formula: radar-v5. This is ReelStack’s calculation, not a YouTube rating or a measure of factual accuracy.
Show the calculation
- 45% performance: 0.64× views vs 7 other channel videos observed at a similar age. View-evidence factor 97% (views / (views + 1,000))
- 25% freshness: 92/100 with a 14-day half-life
- 20% momentum: 14414 views/day vs 24932 channel baseline (7 comparable uploads), with the same view-evidence factor
- 10% engagement: 10/100; likes + 4× comments, smoothed with a 500-view neutral prior