Gareth Howells (Owner of SN85) Reveals Exciting Things About VidaioOS

When Vidaio launched on Bittensor Subnet 85, the pitch was to compress video files smaller, upscale them better, run both through a miner network optimizing on quality scores. That was useful for getting started but turned out to undersell what the product was capable of.

The founder of Vidaio (SN85), Gareth Howells, sat with Gordon Frayne to walk through what the team has built since, and the headline is that Vidaio has evolved from a two-feature pipeline into VidaioOS, an agentic operating system that sits across enterprise video archives and decides what needs to happen to every file in real time.

The Key Points From the Conversation

The substance of what Vidaio has built and where it is heading:

a. Vidaio has transitioned from a single-task pipeline to VidaioOS, an agentic operating system that connects to a client’s archive, reads the contents, and makes recommendations on what each file needs based on what the client is optimizing for.

b. Miners now operate as agentic skill providers rather than single-task workers, with compression and upscaling joined by colorization, content cleanup, localization, and lip-sync.

c. Vidaio compressed files 23% smaller than the FFmpeg benchmark at equivalent VMAF scores, the industry-standard measure of how the human eye perceives video quality.

d. AWS MediaConvert was 97% larger than the FFmpeg benchmark for the same perceptual quality, with Bitmovin sitting between the two.

e. Netflix reportedly spends around $1 billion annually with AWS for infrastructure and storage, which makes a 23% reduction at matched quality a meaningful number on its own.

f. The Pip Studios partnership puts Vidaio one integration step from Netflix, Apple, Amazon, and Paramount, with Pip currently integrating VidaioOS into its existing post-production workflows, particularly around localization and dubbing.

g. The revenue model splits into two paths: one-time archive optimization jobs charged as a percentage of cost savings, and monthly recurring revenue for clients with continuous content ingestion like creator platforms or security operators.

h. Cross-subnet integrations already underway include Hippius (SN75) for storage, Vocence (SN78) for voice cloning, Babelbit (SN59) for localization, Beam (SN105) for streaming, and BitMind (SN34) for AI-generated content detection.

i. Lip-sync is coming as a new competition model, enabling AI-dubbed content to match speakers’ mouth movements in the target language and unlocking monetization of legacy archives in new markets.

j. Gen AI upscaling models in testing take low-resolution legacy content to 4K quality, with early results described as significant.

k. Profits will fund $SN85 buybacks, with the team committed to pumping revenue back into the subnet rather than building a separate external treasury.

Beyond Compression

Vidaio’s evolution from a compression-and-upscaling subnet into VidaioOS reframes what kind of business a Bittensor subnet can structurally become.

The Pip Studios partnership puts Vidaio adjacent to the largest content libraries on the planet, the benchmark numbers give the product a defensible value proposition against AWS and Bitmovin, and the agentic operating model converts discrete tasks into a continuous optimization layer that sits across an entire archive. 

The next test is whether the post-launch sales motion converts the pipeline of conversations into signed enterprise contracts, but the foundation underneath looks sharper than it did when the project first started. Vidaio is positioning itself as the video infrastructure layer for enterprise, and the technical and commercial groundwork to support that claim is now actually in place.