if you tried hyperframes, you would know the agentic video stack is being built on code-gen but SWE-bench code v.s. code-to-video that feels alive are two different things we built the Code2Video Bench, in collab w/ Google DeepMind & @Kaggle frontier labs can finally get good at agentic video tasks

Sep 21, 2026 · 5:17 PM UTC

473
361
1,402
1,665,013
Read the full report here: heygen.com/research/introduc…

Code2Video Benchmark: Evaluating AI Agents on Motion Design | HeyGen Research

HeyGen Research releases the Code2Video Benchmark and a pairwise Judge model to measure how well LLM agents turn prompts into motion graphics code. Leaderboard, failure modes and dataset on Kaggle.

heygen.com
1
4
107
43,668
the dataset and most up-to-date ELO leaderboard on @kaggle: kaggle.com/benchmarks/heygen…

Code2Video Bench Leaderboard | Kaggle

Evaluating Frontier LLMs’ Ability to Create Human-Quality Videos & Motion Graphics from Code

kaggle.com
1
42
24,636
Sort replies: Relevant Recent Liked
Replying to @HyperFrames_ @kaggle
Wait what? GPT-5.5 is the best?
1
4
935
Yes - by a hair and within confidence interval
5
745
Replying to @HyperFrames_ @kaggle
very excited to see Code2Video Bench come to life, awesome work by the Hyperframes / Heygen team!!!
5
74
5,621
Replying to @HyperFrames_ @kaggle
Awesome Work!
4
406
Replying to @HyperFrames_ @kaggle
The benchmark we’ve needed
1
4
301
Replying to @HyperFrames_ @kaggle
Lets go Hyperframes 🚢🚢🚢
2
108
Replying to @HyperFrames_ @kaggle
Exactly Code2Video Bench bridging code gen to video that feels alive is the missing eval we needed!
43
Replying to @HyperFrames_ @kaggle
Kinda surprised this didn’t exist already
101
Replying to @HyperFrames_ @kaggle
This is where things start moving fast
80
Replying to @HyperFrames_ @kaggle
feedback loops win
61
Replying to @HyperFrames_ @kaggle
public leaderboard or it didn't happen
22
Replying to @HyperFrames_ @ExpertWithElora @kaggle
168 different briefs is serious
73
Replying to @HyperFrames_ @kaggle
kinda love how simple that is
52
Replying to @HyperFrames_ @kaggle
The fact labs can train against this is the interesting bit
108
Replying to @HyperFrames_ @kaggle
Love a public leaderboard
25
Replying to @HyperFrames_ @kaggle
long overdue!
1
88
Replying to @HyperFrames_ @kaggle
There’s probably a ton hiding in that data
167
Replying to @HyperFrames_ @kaggle
This is how the quality starts compounding
23
Replying to @HyperFrames_ @kaggle
can't wait to see people arguing over this leaderboard lol
33
Replying to @HyperFrames_ @kaggle
I really like that video needs its own way to test these agents.
45
Replying to @HyperFrames_ @kaggle
video finally has receipts
39
Replying to @HyperFrames_ @kaggle
okay i’d actually browse through these matchups
76
Replying to @HyperFrames_ @kaggle
feels like ai video just got a scoreboard
64
Replying to @HyperFrames_ @kaggle
no more guessing from cherry-picked demos
47
Replying to @HyperFrames_ @kaggle
this feels like a real step forward
85
Replying to @HyperFrames_
168 briefs is commitment
1
112
Replying to @HyperFrames_ @kaggle
this space moves ridiculously fast
1
119
Replying to @HyperFrames_ @kaggle
ai video has a report card now lol
1
1
98
Replying to @HyperFrames_ @kaggle
we need more benchmarks like this
1
101
Replying to @HyperFrames_ @kaggle
the training angle is the sleeper here
24
Replying to @HyperFrames_ @kaggle
finally not an eval you need a phd to understand
138
Replying to @HyperFrames_ @kaggle
this could push video quality forward fast
103
Replying to @HyperFrames_ @kaggle
ai video has needed this badly
38
Replying to @HyperFrames_ @kaggle
Interested to see what builders do with it
111
Load more