README
🚀
AI & ML interests
Pushing the limits of Training Data
Recent Activity
Overlaiapp
updated a
Space 5 days ago
Overlaiapp
posted an update 6 days ago
Post
87
Overlaiai/8K-Resolution
Training Data Maxxxing Month" continues with another very simple question:
Does the model care where the pixels came from?
We just released '8K Resolution', a controlled benchmark with matched source clips in:
• native 8K
• clean 1080p
• synthetic 4K derived from the 1080p version
The goal is not just to ask whether 8K is better.
We want to know:
• Does native spatial detail improve model performance?
• Can models detect upscaled or synthetic resolution?
• Can spatial provenance become a useful filtering signal for scraped video datasets?
• Does a file labeled 4K actually contain native 4K spatial information?
If you test it, find something weird, or build a detector with it, we'd love to see the results.
Training Data Maxxxing Month" continues with another very simple question:
Does the model care where the pixels came from?
We just released '8K Resolution', a controlled benchmark with matched source clips in:
• native 8K
• clean 1080p
• synthetic 4K derived from the 1080p version
The goal is not just to ask whether 8K is better.
We want to know:
• Does native spatial detail improve model performance?
• Can models detect upscaled or synthetic resolution?
• Can spatial provenance become a useful filtering signal for scraped video datasets?
• Does a file labeled 4K actually contain native 4K spatial information?
If you test it, find something weird, or build a detector with it, we'd love to see the results.
Overlaiapp
updated a
dataset 6 days ago
Overlaiapp
published a
dataset 6 days ago
Overlaiapp
updated a
dataset 7 days ago
Overlaiapp
updated a
dataset 8 days ago
Overlaiapp
posted an update 12 days ago
Post
117
Overlaiai/Temporal-Fidelity
"Training Data Maxxxing Month" starts with a simple question:
Does the model care where the video frames came from?
We just released 'Temporal Fidelity', a controlled benchmark with matched source clips in:
• native 60p
• clean 24p
• synthetic 30p derived from the 24p version
The goal is not just to ask whether 60p is better. (it is)
We want to know:
• Does temporal contamination hurt model performance?
• Can models detect temporal junk?
• Can temporal provenance become a useful filtering signal for scraped video datasets?
• Does a file labeled 30 fps actually contain 30 unique captured moments per second?
If you test it, find something weird, or build a detector with it, we'd love to see the results.
"Training Data Maxxxing Month" starts with a simple question:
Does the model care where the video frames came from?
We just released 'Temporal Fidelity', a controlled benchmark with matched source clips in:
• native 60p
• clean 24p
• synthetic 30p derived from the 24p version
The goal is not just to ask whether 60p is better. (it is)
We want to know:
• Does temporal contamination hurt model performance?
• Can models detect temporal junk?
• Can temporal provenance become a useful filtering signal for scraped video datasets?
• Does a file labeled 30 fps actually contain 30 unique captured moments per second?
If you test it, find something weird, or build a detector with it, we'd love to see the results.
Overlaiapp
updated a
dataset 14 days ago
Overlaiapp
published a
dataset 14 days ago
Overlaiapp
posted an update 19 days ago
Post
117
October = Training Data Maxxxing Month.
We’re releasing one controlled video-data test every week in October, free on Hugging Face.
The goal is simple: isolate one variable at a time, then stack them.
Oct 5: 24p vs 30p vs 60p
Oct 12: SDR vs HDR
Oct 19: 1080p vs 4K vs 8K
Oct 26: 8K + 60p + HDR
Same scenes. Controlled comparisons. Real-world footage.
If you train or evaluate video models, we want to see what changes: evals, ablations, side-by-side generations, temporal consistency, training efficiency, or any measurable model improvement.
The point isn’t to assume higher-quality source data is better.
It’s to test it.
Week 4 combines everything into the full Overlai "Gold Standard" stack.
One variable at a time. Then stack them.
#TrainingData #VideoAI #MachineLearning #Datasets #HuggingFace #Overlai
We’re releasing one controlled video-data test every week in October, free on Hugging Face.
The goal is simple: isolate one variable at a time, then stack them.
Oct 5: 24p vs 30p vs 60p
Oct 12: SDR vs HDR
Oct 19: 1080p vs 4K vs 8K
Oct 26: 8K + 60p + HDR
Same scenes. Controlled comparisons. Real-world footage.
If you train or evaluate video models, we want to see what changes: evals, ablations, side-by-side generations, temporal consistency, training efficiency, or any measurable model improvement.
The point isn’t to assume higher-quality source data is better.
It’s to test it.
Week 4 combines everything into the full Overlai "Gold Standard" stack.
One variable at a time. Then stack them.
#TrainingData #VideoAI #MachineLearning #Datasets #HuggingFace #Overlai
LukeNeumann
posted an update almost 2 years ago
Post
1,065
I had a question about Trending datasets. Our initial dataset "Oregon Coast in 4K" was trending at #3 for video at about 700 downloads.
Over the past two days our downloads have spiked, now up to over 2,000, but the dataset has dropped down to the 3rd or 4th page of Trending.
What metrics are used to determine dataset Trending position?
Over the past two days our downloads have spiked, now up to over 2,000, but the dataset has dropped down to the 3rd or 4th page of Trending.
What metrics are used to determine dataset Trending position?
- 1 reply
LukeNeumann
posted an update almost 2 years ago
Post
584
Something I've been thinking about regarding fine-tuning video models.
How well are these models trained on the fundamentals of image capture? Take shutter speed/angle, for instance.
The difference in a video with a 1/2000 speed shutter and a 1/60 speed shutter is drastic.
What about color science? HDR vs. SDR, Anamorphic vs. spherical, sensor size, aspect ratio.
Is it worth doing an open-source dataset "film school series"?
It may seem too granular but I wonder if this level of granularity makes for better downstream results.
How well are these models trained on the fundamentals of image capture? Take shutter speed/angle, for instance.
The difference in a video with a 1/2000 speed shutter and a 1/60 speed shutter is drastic.
What about color science? HDR vs. SDR, Anamorphic vs. spherical, sensor size, aspect ratio.
Is it worth doing an open-source dataset "film school series"?
It may seem too granular but I wonder if this level of granularity makes for better downstream results.
- 1 reply
LukeNeumann
posted an update almost 2 years ago
Post
1,257
Nine years ago, I uploaded the first 8K resolution video to YouTube and I've been stockpiling 8K footage ever since: https://www.youtube.com/watch?v=sLprVF6d7Ug&t
Should @Overlaiapp release the first open-source 8K video dataset?
Could anyone even fine tune a model with this?😅
Should @Overlaiapp release the first open-source 8K video dataset?
Could anyone even fine tune a model with this?😅
- 6 replies
demid-overlai
updated a
dataset almost 2 years ago
LukeNeumann
posted an update almost 2 years ago
Post
1,883
Hello Hugging Face community!
I wanted to introduce myself and my company @Overlaiapp . We are a collective of filmmakers, photographers, and AI engineers working on high resolution (8K+) training data.
We plan to share a lot of our datasets with the community and are kicking things off with two curated datasets:
- Overlaiai/OregonCoastin4K
- Overlaiai/SubArcticPolarBear
Overlai.ai Dataset Features
🎥 Oversampled: Every clip is captured in stunning 8K resolution, delivering rich detail ideal for fine tuning scenic landscapes and ocean dynamics.
📸 Variance: Includes close-up details, slow-motion footage of crashing waves, sweeping landscapes, and wildlife shots.
📋 Detailed Metadata: Every clip is paired with structured metadata, including creative descriptions, precise camera movements, lens information, field of view calculations, and shot settings, ensuring AI models can fully understand and replicate real-world cinematography with accuracy.
⚙️ Consistency: Re-thinking training data at the point of capture by "overshooting" a subject, enabling models to learn more nuanced relationships and views across scenes.
🌅 Light: Shot during early morning and sunset light for optimal color contrast and dynamic range, maximizing visual quality for color and lighting-sensitive tasks.
🔍 Curation: Curated specifically for machine learning, providing clean, high-quality data for next generation model training.
I wanted to introduce myself and my company @Overlaiapp . We are a collective of filmmakers, photographers, and AI engineers working on high resolution (8K+) training data.
We plan to share a lot of our datasets with the community and are kicking things off with two curated datasets:
- Overlaiai/OregonCoastin4K
- Overlaiai/SubArcticPolarBear
Overlai.ai Dataset Features
🎥 Oversampled: Every clip is captured in stunning 8K resolution, delivering rich detail ideal for fine tuning scenic landscapes and ocean dynamics.
📸 Variance: Includes close-up details, slow-motion footage of crashing waves, sweeping landscapes, and wildlife shots.
📋 Detailed Metadata: Every clip is paired with structured metadata, including creative descriptions, precise camera movements, lens information, field of view calculations, and shot settings, ensuring AI models can fully understand and replicate real-world cinematography with accuracy.
⚙️ Consistency: Re-thinking training data at the point of capture by "overshooting" a subject, enabling models to learn more nuanced relationships and views across scenes.
🌅 Light: Shot during early morning and sunset light for optimal color contrast and dynamic range, maximizing visual quality for color and lighting-sensitive tasks.
🔍 Curation: Curated specifically for machine learning, providing clean, high-quality data for next generation model training.
demid-overlai
updated a
dataset almost 2 years ago
Chaseteron
updated a
dataset almost 2 years ago
Update README.md
#2 opened almost 2 years ago
by 
Chaseteron
Upload header (2).jpg
#1 opened almost 2 years ago
by Chaseteron
Update README.md
#1 opened almost 2 years ago
by demid-overlai