Video thumbnails are often the first things viewers see when they look for something interesting to watch. A strong, vibrant, and relevant thumbnail draws attention, giving viewers a quick preview of the content of the video, and helps them to find content more easily. Better thumbnails lead to more clicks and views for video creators.

Inspired by the recent remarkable advances of deep neural networks (DNNs) in computer vision, such as image and video classification, our team has recently launched an improved automatic YouTube "thumbnailer" in order to help creators showcase their video content. Here is how it works.

The Thumbnailer Pipeline
While a video is being uploaded to YouTube, we first sample frames from the video at one frame per second. Each sampled frame is evaluated by a quality model and assigned a single quality score. The frames with the highest scores are selected, enhanced and rendered as thumbnails with different sizes and aspect ratios. Among all the components, the quality model is the most critical and turned out to be the most challenging to develop. In the latest version of the thumbnailer algorithm, we used a DNN for the quality model. So, what is the quality model measuring, and how is the score calculated?

The main processing pipeline of the thumbnailer.

(Training) The Quality Model
Unlike the task of identifying if a video contains your favorite animal, judging the visual quality of a video frame can be very subjective - people often have very different opinions and preferences when selecting frames as video thumbnails. One of the main challenges we faced was how to collect a large set of well-annotated training examples to feed into our neural network. Fortunately, on YouTube, in addition to having algorithmically generated thumbnails, many YouTube videos also come with carefully designed custom thumbnails uploaded by creators. Those thumbnails are typically well framed, in-focus, and center on a specific subject (e.g. the main character in the video). We consider these custom thumbnails from popular videos as positive (high-quality) examples, and randomly selected video frames as negative (low-quality) examples. Some examples of the training images are shown below.

Example training images.

Results
Compared to the previous automatically generated thumbnails, the DNN-powered model is able to select frames with much better quality. In a human evaluation, the thumbnails produced by our new models are preferred to those from the previous thumbnailer in more than 65% of side-by-side ratings. Here are some examples of how the new quality model performs on YouTube videos:

Example frames with low and high quality score from the DNN quality model, from video “Grand Canyon Rock Squirrel”.

Thumbnails generated by old vs. new thumbnailer algorithm.

We recently launched this new thumbnailer across YouTube, which means creators can start to choose from higher quality thumbnails generated by our new thumbnailer. Next time you see an awesome YouTube thumbnail, don’t hesitate to give it a thumbs up. ;)

How to start developing

• Choose your reports:
• Video reports provide statistics for all user activity related to a channel's videos or a content owner's videos. For example, these metrics include the number of views or ratings that videos received. Some video reports for content owners also include earnings and ad performance metrics.
• Playlist reports provide statistics that are specifically related to video views that occur in the context of a playlist.
• Ad performance reports provide impression-based metrics for ads that ran during video playbacks. These metrics account for each ad impression, and each video playback can yield multiple impressions.
• Estimated earnings reports provide the total earnings for videos from Google-sold advertising sources as well as from non-advertising sources. These reports also contain some ad performance metrics.
• Asset reports provide user activity metrics related to videos that are linked to a content owners' assets. For its data to included in the report, a video must have been uploaded by the content owner and then claimed as a match of an asset in the YouTube Content ID system.

• Schedule reports:

1. Get an OAuth token (authentication credentials)
2. Call the reportTypes.list method to retrieve a list of the available report types
3. Create a new reporting job by calling jobs.create and passing the desired report type (and/or query in the future)

• Retrieve reports:

1. Get an OAuth token (authentication credentials)
2. Call the jobs.list method to retrieve a list of the available reporting jobs and remember its ID.
3. Call the reports.list method with the jobId filter parameter set to the ID found in the previous step to retrieve a list of downloadable reports that that particular job created.
4. Creators can check the report’s last modified date to determine whether the report has been updated since the last time it was retrieved.
5. Fetch the report from the URL obtained by step 3.

• While using our sample code and tools
• Client libraries for many different programming languages can help you implement the YouTube Reporting API as well as many other Google APIs.
• Don't write code from scratch! Our Java, PHP, and Python code samples will help you get started.
• The APIs Explorer lets you try out sample calls before writing any code.

2005: YouTube is born

Me at the Zoo is the first video uploaded to YouTube

2006: Google buys YouTube

One year after YouTube launches, videos play in the FLV container with the H.263 codec at a maximum resolution of 240p. We scale videos up to 640x360, but you can still click a button to play at original size.

2007: YouTube goes mobile

YouTube is one of the original applications on the iPhone. Because it doesn't support Flash, we re-encode every single YouTube video into H.264 with the MP4 container. YouTube videos get a resolution notch to 360p.

2008: YouTube kicks it up to HD

With upload sizes and download speeds growing, videos jump in size up to 720p HD. Lower resolution files get higher quality by squeezing Main Profile H.264 into FLVs.

2009: YouTube enters the third dimension

YouTube supports 3D videos, 1080p and live streaming.

2010: YouTube's on TV

The biggest screen in your house now gets YouTube courtesy of Flash Lite and ActionScript 2. 2010 also sees the first playbacks with HTML5 <video> thanks to VP8, an open source video codec. We bump up the maximum resolution to 4K, known as "Original" at the time.

2011: YouTube slices bread (and videos) to battle buffering

We launch Sliced Bread, codename for a project that enables adaptive bitrate in the Flash player by requesting videos a little piece at a time. Users see higher quality videos more often and buffering less often.

2012: YouTube live streaming hits prime time

We scale up our live streaming infrastructure to support the 2012 Summer Olympics, with over 1,200 events. In October, over 8 million people watch live as Felix Baumgartner jumps from the stratosphere.

2013: YouTube's first taste of VP9

We start our first experiments with VP9 in Chrome, which brings higher quality video at less bandwidth. Adaptive bitrate streaming in the HTML5 and Flash players moves to the DASH standard using both FMP4 and MKV video containers.

2014: Silky smooth 60fps comes to YouTube

High frame rate isn't just for games anymore: YouTube now supports videos that play in up to 60fps. Gangnam Style becomes the first YouTube video to break the MAX_INT barrier with more than 2³² / 2 - 1 views.

2015: YouTube adds spherical video (look behind you!)

You can now upload videos that wrap 360 degrees around the viewer. Even 4K videos can play up to 60fps. HTML5 becomes the default YouTube web player.

Richard Leider, Engineering Manager, recently watched David Bowie - Oh You Pretty Things
Jonathan Levine, Product Manager, recently watched Candide Thovex - One of those days 2

As more people watch more high-quality videos across more screens, we need video formats that provide better resolution without increasing bandwidth usage. That’s why we started encoding YouTube videos in VP9, the open-source codec that brings HD and even 4K (2160p) quality at half the bandwidth used by other known codecs.

VP9 is the most efficient video compression codec in widespread use today. In the last year alone, YouTube users have already watched more than 25 billion hours of VP9 video, billions of which would not have been played in HD without VP9's bandwidth benefits. And with more of our device partners adopting VP9, we wanted to give you a primer on the technology.

How VP9 works

Videos hold a lot of information. If video were stored in the same format that a camera sensor uses when shooting a scene, the resulting files would be enormous — raw 4K is up to 18,000 Mbps! Instead, modern video compression looks at a video more like a person might, by encoding a description of the features in a scene, and tracking how those features move and change. This compression is hundreds of times more efficient than a camera sensor's recording and is what makes video streaming possible.

While VP9 uses the same basic blueprint as previous codecs, the WebM team has packed improvements into VP9 to get more quality out of each byte of video. For instance, the encoder prioritizes the sharpest image features, and the codec now uses asymmetric transforms to help keep even the most challenging scenes looking crisp and block-free.

Here's a comparison between the image quality you'd get watching Janelle Monaé with VP9 or legacy H.264 transcodes on a 600Kbps connection:

Bringing quality to the people

This new format bumps everybody one notch closer to our goal of instant, high-quality, buffer-free videos. That means that if your Internet connection used to only play up to 480p without buffering on YouTube, it can now play silky smooth 720p with VP9.

VP9 also has benefits for people with limited bandwidth or expensive data plans. By cutting bitrates in as much as half, it dramatically increases the set of users that can watch 360p quality video without increased rebuffering or cost.

Opening the door to 4K

And for those who can never get enough pixels (including your humble author!), VP9 unlocks the burgeoning world of 4K videos. At larger video sizes, VP9 actually gets even more efficient than its predecessors, so uninterrupted 4K content can now be streamed by a significant and growing part of the YouTube audience. The amount of 4K video uploaded to YouTube has more than tripled in the past year, and VP9 helps us plan for improved streaming into the future. You can find 4K videos by using the search filter, or see some of our favorites in this playlist.

Where can I use VP9?

Thanks to our device partners, VP9 decoding support is available today in the Chrome web browser, in Android devices like the Samsung Galaxy S6, and in TVs and game consoles from Sony, LG, Sharp, and more. More than 20 device partners across the industry are launching products in 2015 and beyond using VP9.

To learn more about producing your own VP9 content, see our FFMpeg encoding guide or check out the Adobe Premier WebM plugin.

Steven Robertson, Software Engineer, recently watched “St. Lucia - Before The Dive.”

• Add more replicas for better read throughput and data durability
• Introduce sharding to scale your write throughput and let your data set grow beyond a single machine
• Create separate replica pools for batch jobs and backups, to isolate them from live traffic
• Clone the whole deployment into multiple datacenters worldwide for disaster recovery and lower latency

• Horizontal scaling – add capacity by launching additional nodes rather than making one huge node
• Dynamic placement – let the cluster manager schedule Vitess containers wherever it wants
• Declarative specification – describe your desired end state, and let the cluster manager create it
• Self-healing components – recover automatically from machine failures