August was mostly about foundations.
BaseTube is still in beta, and before we open it to more creators, a few parts of the platform need to be solid: uploading, media delivery, search, content management, and analytics.
Over the past few weeks, we shipped major improvements across all five.
Creators already have platforms for discovery. YouTube, TikTok, Instagram and others are very good at distributing attention.
What creators have much less control over is what happens after that attention exists: where their content lives, what they know about their audience, how they monetize it, and how much of that relationship they actually own.
That is the layer we are building BaseTube around.
Before we can build deeper monetization and audience tools on top, the underlying video platform has to be reliable enough to support them.
This month moved us significantly closer.
Uploads are now faster, resumable, and much harder to break
Video uploads no longer pass through the BaseTube application servers.
Files now upload directly from the browser in resumable chunks, so an interrupted connection does not mean starting again from zero. Close the tab, lose Wi-Fi, or come back later and the upload can continue from where it stopped.
Creators can also queue multiple videos at once from Content Studio and continue working while files upload and process.
Behind the scenes, upload, inspection, transcoding, indexing and cleanup are handled independently. That gives us a cleaner processing pipeline and more room to scale without tying video processing to the web application itself.
The current upload path uses Cloudflare R2 as a fast ingestion layer, while completed video assets continue to be transferred into Storj, which remains BaseTube’s long-term storage layer.
We also avoid transcoding files unnecessarily. If a video is already in a format browsers can reliably play, BaseTube can publish it without sending it through a full encoding cycle.
For most standard MP4 uploads, that means much faster publishing.
A 1.5 GB video was one of our test cases, not the system limit.
Search is now instant
BaseTube search now runs on Meilisearch.
Search supports typo tolerance, highlighted matches, category and duration filters, synonyms, and suggestions while typing.
A search for marakech, for example, can still find Marrakech.
Newly published videos are indexed within seconds instead of waiting for a scheduled job.
We have also kept a database fallback behind the search layer. If the search service becomes unavailable, BaseTube can continue returning results in degraded mode rather than simply failing.
For creators with hundreds or thousands of videos, search is not cosmetic. It is part of how the catalogue remains usable over time.
Faster media delivery
Thumbnails, posters and other frequently requested media assets can now be cached and served closer to the viewer instead of every request travelling back to the original storage layer.
The result is faster and more predictable page loading, particularly for audiences farther from the underlying infrastructure.
This is a delivery optimisation, not a storage migration. Storj remains the durable storage layer for completed video assets.
Content Studio is becoming a proper creator workspace
A large part of the work this month also went into making the Creator Hub more useful day to day.
Video management now includes inline Public/Private controls, undo support, bulk actions, search, filtering and sorting.
Those filters live in the URL, so a specific working view can be bookmarked or shared.
The editing experience is also tighter, with ⌘S saving, protection against leaving with unsaved changes, and a more compact interface.
We have also continued integrating AI where it removes repetitive work rather than simply adding AI for the sake of it.
AI-generated descriptions now produce properly structured video copy instead of generic text blocks, and AI thumbnail generation is integrated directly into the Creator Hub with its own quota system.
Individually, these are small improvements.
Together, they are starting to turn Content Studio from an administration page into the workspace we eventually want creators operating from.
Analytics built around data we can actually measure
Creator analytics received some of the most important work this month.
Our rule is simple:
If BaseTube cannot measure something reliably, we would rather not show it than display a precise-looking number we cannot defend.
Watch time is now based on time actually played.
Playback is measured using regular heartbeats, with additional events when a viewer pauses, closes the page or leaves the tab. Scrubbing through a video does not count as watch time, and mobile playback is now captured correctly.
Average percentage watched is calculated per view. Engagement cards use real period totals. Trend comparisons now compare equivalent windows.
And when data cannot be loaded reliably, the dashboard shows an error state instead of quietly presenting a zero.
This foundation also changes what we can do with AI Insights.
The goal is not to generate generic advice such as “make better thumbnails.”
AI Insights can analyse a creator’s own catalogue across titles, thumbnails and positioning, identify gaps tied to specific videos, prioritise changes and provide per-video observations.
Creators can also compare their channel positioning against the wider YouTube niche using a simple structure:
They do.
You do.
Try next.
Performance-based conclusions only appear when there is enough actual viewing data to support them.
Any numerical claim shown by the AI has to be traceable back to the underlying analytics.
AI is very good at producing confident explanations. That does not mean those explanations should automatically appear inside a creator analytics product.
Why we’re building these foundations
It would be easy to look at this release as uploads, search, delivery, and dashboards.
For us, these are prerequisites.
BaseTube is being built around a broader idea: creators should be able to convert attention they generate elsewhere into a relationship they control directly.
That means knowing who supports their work.
Knowing what those people watch.
Being able to sell access to content.
Understanding what converts.
Communicating directly with the audience around that content.
And having more control over the economics of that relationship.
Content Passes are already one part of that model.
The creator and audience tooling around them is where much of the next phase of BaseTube is heading.
Before those layers become genuinely valuable, BaseTube has to be able to ingest content reliably, organise it, deliver it globally and measure what happens around it accurately.
That is what August was largely about.
There is still a lot to build.
But the difference between the BaseTube beta at the beginning of August and the platform today is substantial.
The foundations are getting into place.
Now we can increasingly focus on what we wanted to build on top of them in the first place.
BaseTube is currently available in beta at beta.base.tube.