Subnet Directory

With the amount of new subnets being added it can be hard to get up to date information across all subnets, so data may be slightly out of date from time to time

Subnet 85

Vidaio

Emissions
Value
Recycled
Value
Recycled (24h)
Value
Registration Cost
Value
Active Validators
Value
Active Miners
Value
Active Dual Miners/Validators
Value

ABOUT

What exactly does it do?

Vidaio is an open-source video processing subnet focused on AI-driven video upscaling (and soon compression/streaming). Its mission is to “democratise video enhancement through decentralisation, artificial intelligence, and blockchain technology”, providing creators and businesses with scalable, affordable, high-quality video processing. The Vidaio team emphasizes a decentralized, community-led approach: the project’s code is fully open-source and built atop Bittensor’s merit-based model. By leveraging Bittensor, Vidaio can tap into a distributed pool of GPUs and reward contributors via on-chain incentives, ensuring continual improvement of its AI models.

Vidaio uses a number of deep learning models to enhance low-resolution videos by predicting and generating high-resolution details. Unlike traditional upscaling methods, which rely on interpolation, their AI-based techniques analyze patterns, textures, and edges to reconstruct sharper and more realistic frames. Vidaio can instantly adapt to growing demand making it a viable solution for businesses of all sizes and eliminates a single point of control, enhancing security and censorship resistance.

Vidaio is an open-source video processing subnet focused on AI-driven video upscaling (and soon compression/streaming). Its mission is to “democratise video enhancement through decentralisation, artificial intelligence, and blockchain technology”, providing creators and businesses with scalable, affordable, high-quality video processing. The Vidaio team emphasizes a decentralized, community-led approach: the project’s code is fully open-source and built atop Bittensor’s merit-based model. By leveraging Bittensor, Vidaio can tap into a distributed pool of GPUs and reward contributors via on-chain incentives, ensuring continual improvement of its AI models.

Vidaio uses a number of deep learning models to enhance low-resolution videos by predicting and generating high-resolution details. Unlike traditional upscaling methods, which rely on interpolation, their AI-based techniques analyze patterns, textures, and edges to reconstruct sharper and more realistic frames. Vidaio can instantly adapt to growing demand making it a viable solution for businesses of all sizes and eliminates a single point of control, enhancing security and censorship resistance.

PURPOSE

What exactly is the 'product/build'?

Vidaio follows the standard Bittensor subnet design with miners and validators cooperating via two “synapses” (query types).

  • Miners: Run AI upscaling models on video data. Miners “enhance video quality using AI-driven upscaling techniques.” They may fine-tune open-source super-resolution models or develop new ones, and they handle upscaling requests from validators and end-users. In practice, miners take low-resolution video chunks (from synthetic or organic queries) and output high-resolution versions.
  • Validators: Check that miners’ outputs meet quality standards. Validators feed controlled test videos to miners and “ensure miners deliver consistent, high-quality results by evaluating performance”. In a Synthetic Query, validators downscale a known high-resolution clip, send it to miners to upscale, then score the result using perceptual metrics (Vidaio uses VMAF and PIE-APP for this). In an Organic Query, real user videos are split into chunks and queued to miners; once upscaled, the pieces are reassembled and returned to the user. Validators then use a no-reference quality metric (Vidaio uses ClipIQA+ for organic jobs) to assess the output’s perceptual quality.
  • Incentive Mechanism: Vidaio’s reward system is built on video-quality metrics. Validators periodically submit weight vectors reflecting each miner’s scores on the subnet. The Bittensor blockchain aggregates these vectors and applies the Yuma consensus algorithm, which combines each miner’s weighted score with its staked TAO to compute token emissions. Thus, miners earn more TAO by consistently producing high-VMAF/PIE-APP scores, and validators are themselves rewarded for honest scoring. This merit-based model (unique to each subnet) drives the network to optimize for perceptual video quality. (In benchmark tests, Vidaio’s upscaler outperformed leading centralized solutions – e.g. its ClipIQA+ score was 0.4697 vs. 0.4658 for Topaz – demonstrating strong perceptual performance.)

 

Products & Services

The product is a decentralized video processing solution offered through the Bit Tensor network, which is designed to address the increasing demand for high-quality, low-cost video processing tools. This subnet provides two core services:

AI-Driven Video Upscaling: It enhances lower-resolution video files (e.g., SD to HD or 4K) using machine learning models that predict the correct pixel values for higher resolution output. This allows old or low-quality content (such as SD footage from the ’90s) to be transformed into HD or even 4K video, making it more suitable for modern devices and platforms. This service is already live in a beta phase and has shown great success with positive community feedback.

AI-Driven Video Compression: This service focuses on reducing video file sizes without compromising visual quality. The compression model analyzes the content, applies optimized encoding, and eliminates redundancies to achieve significant reductions in file size (up to 80% smaller), while maintaining the same perceptual quality. Unlike traditional methods, the model adapts the compression process based on the unique characteristics of the video (e.g., dynamic motion or static scenes) to achieve optimal results. This is set to revolutionize video delivery, especially in content-heavy industries like streaming, by drastically lowering storage and bandwidth costs.

 

Technical Architecture

Vidaio’s tech stack combines decentralized blockchain components with AI/ML video pipelines. Core elements include:

  • Bittensor Integration: The subnet runs on the Bittensor chain. Both miners and validators use the Bittensor Python SDK/CLI to register, stake, and communicate on-chain. Validators regularly send weight vectors to the chain, and the Bittensor network’s Yuma consensus module computes emissions. (This is standard for all Bittensor subnets.) Vidaio follows the dynamic TAO model, meaning TAO tokens staked into Vidaio (via the Bittensor wallet) generate dTAO reflecting Vidaio’s value and share in emissions.
  • AI/ML Models: Vidaio’s upscaling engine uses state-of-the-art super-resolution models. The code references known open-source frameworks (e.g. Video2X, ESRGAN variants) and implements novel enhancements. Training and inference are done in Python (likely using PyTorch or TensorFlow libraries, given the requirements). The repository’s Docker setup suggests containerized deployment for reproducibility. For quality evaluation, Vidaio employs multiple metrics (VMAF, PIE-APP, TOPIQ, LPIPS, ClipIQA+, etc.) as shown in their documentation.
  • Infrastructure: Although decentralized, typical miner nodes are GPU-equipped servers/instances running Vidaio’s Python mining software. The GitHub structure (folders like vidaio_subnet_core, services, docker) indicates a modular backend. The public web service (vidaio.io) likely runs on cloud servers using this core code plus a web frontend (React/Vue, since a frontend dev is on the team).
  • Interoperability & Tooling: Vidaio interoperates seamlessly with the Bittensor network: it observes the metagraph of peers and follows the Bittensor protocol. It uses standard crypto wallets (multisig hotkeys) for TAO. Tools like Docker, the Bittensor CLI, and monitoring dashboards are part of the stack. The GitHub repo includes setup guides for validators and miners, as well as an “Incentive Mechanism Guide” (linked in the docs) to explain the scoring algorithm.

 

Vidaio follows the standard Bittensor subnet design with miners and validators cooperating via two “synapses” (query types).

  • Miners: Run AI upscaling models on video data. Miners “enhance video quality using AI-driven upscaling techniques.” They may fine-tune open-source super-resolution models or develop new ones, and they handle upscaling requests from validators and end-users. In practice, miners take low-resolution video chunks (from synthetic or organic queries) and output high-resolution versions.
  • Validators: Check that miners’ outputs meet quality standards. Validators feed controlled test videos to miners and “ensure miners deliver consistent, high-quality results by evaluating performance”. In a Synthetic Query, validators downscale a known high-resolution clip, send it to miners to upscale, then score the result using perceptual metrics (Vidaio uses VMAF and PIE-APP for this). In an Organic Query, real user videos are split into chunks and queued to miners; once upscaled, the pieces are reassembled and returned to the user. Validators then use a no-reference quality metric (Vidaio uses ClipIQA+ for organic jobs) to assess the output’s perceptual quality.
  • Incentive Mechanism: Vidaio’s reward system is built on video-quality metrics. Validators periodically submit weight vectors reflecting each miner’s scores on the subnet. The Bittensor blockchain aggregates these vectors and applies the Yuma consensus algorithm, which combines each miner’s weighted score with its staked TAO to compute token emissions. Thus, miners earn more TAO by consistently producing high-VMAF/PIE-APP scores, and validators are themselves rewarded for honest scoring. This merit-based model (unique to each subnet) drives the network to optimize for perceptual video quality. (In benchmark tests, Vidaio’s upscaler outperformed leading centralized solutions – e.g. its ClipIQA+ score was 0.4697 vs. 0.4658 for Topaz – demonstrating strong perceptual performance.)

 

Products & Services

The product is a decentralized video processing solution offered through the Bit Tensor network, which is designed to address the increasing demand for high-quality, low-cost video processing tools. This subnet provides two core services:

AI-Driven Video Upscaling: It enhances lower-resolution video files (e.g., SD to HD or 4K) using machine learning models that predict the correct pixel values for higher resolution output. This allows old or low-quality content (such as SD footage from the ’90s) to be transformed into HD or even 4K video, making it more suitable for modern devices and platforms. This service is already live in a beta phase and has shown great success with positive community feedback.

AI-Driven Video Compression: This service focuses on reducing video file sizes without compromising visual quality. The compression model analyzes the content, applies optimized encoding, and eliminates redundancies to achieve significant reductions in file size (up to 80% smaller), while maintaining the same perceptual quality. Unlike traditional methods, the model adapts the compression process based on the unique characteristics of the video (e.g., dynamic motion or static scenes) to achieve optimal results. This is set to revolutionize video delivery, especially in content-heavy industries like streaming, by drastically lowering storage and bandwidth costs.

 

Technical Architecture

Vidaio’s tech stack combines decentralized blockchain components with AI/ML video pipelines. Core elements include:

  • Bittensor Integration: The subnet runs on the Bittensor chain. Both miners and validators use the Bittensor Python SDK/CLI to register, stake, and communicate on-chain. Validators regularly send weight vectors to the chain, and the Bittensor network’s Yuma consensus module computes emissions. (This is standard for all Bittensor subnets.) Vidaio follows the dynamic TAO model, meaning TAO tokens staked into Vidaio (via the Bittensor wallet) generate dTAO reflecting Vidaio’s value and share in emissions.
  • AI/ML Models: Vidaio’s upscaling engine uses state-of-the-art super-resolution models. The code references known open-source frameworks (e.g. Video2X, ESRGAN variants) and implements novel enhancements. Training and inference are done in Python (likely using PyTorch or TensorFlow libraries, given the requirements). The repository’s Docker setup suggests containerized deployment for reproducibility. For quality evaluation, Vidaio employs multiple metrics (VMAF, PIE-APP, TOPIQ, LPIPS, ClipIQA+, etc.) as shown in their documentation.
  • Infrastructure: Although decentralized, typical miner nodes are GPU-equipped servers/instances running Vidaio’s Python mining software. The GitHub structure (folders like vidaio_subnet_core, services, docker) indicates a modular backend. The public web service (vidaio.io) likely runs on cloud servers using this core code plus a web frontend (React/Vue, since a frontend dev is on the team).
  • Interoperability & Tooling: Vidaio interoperates seamlessly with the Bittensor network: it observes the metagraph of peers and follows the Bittensor protocol. It uses standard crypto wallets (multisig hotkeys) for TAO. Tools like Docker, the Bittensor CLI, and monitoring dashboards are part of the stack. The GitHub repo includes setup guides for validators and miners, as well as an “Incentive Mechanism Guide” (linked in the docs) to explain the scoring algorithm.

 

WHO

Team Info

Vidaio is developed by a multi-disciplinary team of industry professionals (names as listed on the website). Key members include:

Gareth Howells – Founder. Gareth has 20 years of experience in the product and operations field, having worked with major companies like Netflix, Disney, Sony, Spotify, Hulu, and even Pokémon. His expertise lies in understanding market needs and product development. Gareth is particularly focused on taking this product to market and solving real-world problems for content owners and video platforms.

Ahmad Ayad – Machine Learning Engineer (AI/ML specialist). Ahmed is the machine learning expert who primarily focuses on the video compression model. With a PhD in distributed learning, he has been with the team since its inception. Ahmed’s work involves using machine learning models to enhance the efficiency of video compression and apply pre-processing to reduce video artifacts before compression, ensuring better visual quality. His work in video compression has been groundbreaking and forms the core of the product’s capabilities.

Gopi Jayaraman – Video Technology Expert (AV/multimedia veteran).

Medfil D – Subnet Developer (backend & AI/ML engineer).

Chinaza – UI/UX Designer.

Akinwunmi Aguda – Frontend Developer.

Marcus “mogmachine” Graichen – Angel Advisor (cryptoinvestor and Bittensor community figure).

Many of the above contributors have public profiles (e.g. Marcus Graichen is known as “mogmachine” on Bittensor forums/X), and the GitHub repo shows activity by the Vidaio org. The project maintains an active presence on social media and forums: the official X (Twitter) handle is @vidaio_τ (declared as “open-source, decentralized video processing, subnet 85 on Bittensor” in the bio), and a community Discord is open via the invite on their site. All code and documentation are published on GitHub (confirmed by project links on their Medium blog), and the team engages with the Bittensor community for announcements and support.

 

Vidaio is developed by a multi-disciplinary team of industry professionals (names as listed on the website). Key members include:

Gareth Howells – Founder. Gareth has 20 years of experience in the product and operations field, having worked with major companies like Netflix, Disney, Sony, Spotify, Hulu, and even Pokémon. His expertise lies in understanding market needs and product development. Gareth is particularly focused on taking this product to market and solving real-world problems for content owners and video platforms.

Ahmad Ayad – Machine Learning Engineer (AI/ML specialist). Ahmed is the machine learning expert who primarily focuses on the video compression model. With a PhD in distributed learning, he has been with the team since its inception. Ahmed’s work involves using machine learning models to enhance the efficiency of video compression and apply pre-processing to reduce video artifacts before compression, ensuring better visual quality. His work in video compression has been groundbreaking and forms the core of the product’s capabilities.

Gopi Jayaraman – Video Technology Expert (AV/multimedia veteran).

Medfil D – Subnet Developer (backend & AI/ML engineer).

Chinaza – UI/UX Designer.

Akinwunmi Aguda – Frontend Developer.

Marcus “mogmachine” Graichen – Angel Advisor (cryptoinvestor and Bittensor community figure).

Many of the above contributors have public profiles (e.g. Marcus Graichen is known as “mogmachine” on Bittensor forums/X), and the GitHub repo shows activity by the Vidaio org. The project maintains an active presence on social media and forums: the official X (Twitter) handle is @vidaio_τ (declared as “open-source, decentralized video processing, subnet 85 on Bittensor” in the bio), and a community Discord is open via the invite on their site. All code and documentation are published on GitHub (confirmed by project links on their Medium blog), and the team engages with the Bittensor community for announcements and support.

 

FUTURE

Roadmap

The roadmap for the subnet outlines the key stages of development, focusing on scaling and enhancing its capabilities over time.

Upscaling: Currently live in beta, the upscaling feature allows content owners and video platforms to send low-resolution videos to miners for enhancement. This service is actively being tested and refined, with the team using industry-standard metrics (like VMAF and PI app score) to measure the perceptual quality of the upscaled videos. The upscaling service has already gained significant positive feedback from the community, though the team is working to improve the quality further.

 

Video Compression: The compression model is expected to be launched soon, and it promises to revolutionize video storage and bandwidth consumption. By utilizing advanced AI, the compression service can reduce video file sizes by up to 80% without sacrificing perceptual quality. This will have a significant impact on industries that handle massive amounts of video data, including content delivery networks (CDNs) and video on demand (VOD) platforms.

 

Future Developments:

  • On-Demand Video Architecture: The team plans to integrate on-demand video processing for platforms that need to dynamically process and deliver videos to users. This would involve decentralized storage solutions and a decentralized CDN to ensure efficient and cost-effective video delivery.
  • Live Streaming: This will be a more complex development but is part of the long-term vision for the subnet. The goal is to create a decentralized streaming service using the Bit Tensor network, which could drastically reduce streaming costs and increase accessibility.
  • Collaborations: The team is exploring partnerships with other Bit Tensor subnets, such as Hippius (for decentralized storage) and BitMind (for AI services). These collaborations will enhance the overall product offering by integrating AI-powered audio restoration and other capabilities into video processing.

 

Go-to-Market Strategy:

  • Focus on Enterprise Clients: The team is targeting large-scale content owners and streaming platforms as primary clients, as they stand to benefit most from the reduced storage and bandwidth costs that the subnet provides. The solution is particularly valuable for platforms operating in emerging markets, where internet bandwidth is limited, and the need for efficient video compression is critical.
  • Global Expansion: The team aims to expand its reach, starting with Africa, where there is a high demand for efficient video processing solutions. Once the model is proven in this market, it will be adapted for other regions globally.

 

The video subnet is positioned to tackle the $400 billion market for video processing, which includes sectors like streaming, cloud storage, and video post-production. The technology will be used in various industries such as surveillance, medical imaging, and industrial video applications. This makes it a versatile solution with the potential to scale across a broad range of markets.

 

The roadmap for the subnet outlines the key stages of development, focusing on scaling and enhancing its capabilities over time.

Upscaling: Currently live in beta, the upscaling feature allows content owners and video platforms to send low-resolution videos to miners for enhancement. This service is actively being tested and refined, with the team using industry-standard metrics (like VMAF and PI app score) to measure the perceptual quality of the upscaled videos. The upscaling service has already gained significant positive feedback from the community, though the team is working to improve the quality further.

 

Video Compression: The compression model is expected to be launched soon, and it promises to revolutionize video storage and bandwidth consumption. By utilizing advanced AI, the compression service can reduce video file sizes by up to 80% without sacrificing perceptual quality. This will have a significant impact on industries that handle massive amounts of video data, including content delivery networks (CDNs) and video on demand (VOD) platforms.

 

Future Developments:

  • On-Demand Video Architecture: The team plans to integrate on-demand video processing for platforms that need to dynamically process and deliver videos to users. This would involve decentralized storage solutions and a decentralized CDN to ensure efficient and cost-effective video delivery.
  • Live Streaming: This will be a more complex development but is part of the long-term vision for the subnet. The goal is to create a decentralized streaming service using the Bit Tensor network, which could drastically reduce streaming costs and increase accessibility.
  • Collaborations: The team is exploring partnerships with other Bit Tensor subnets, such as Hippius (for decentralized storage) and BitMind (for AI services). These collaborations will enhance the overall product offering by integrating AI-powered audio restoration and other capabilities into video processing.

 

Go-to-Market Strategy:

  • Focus on Enterprise Clients: The team is targeting large-scale content owners and streaming platforms as primary clients, as they stand to benefit most from the reduced storage and bandwidth costs that the subnet provides. The solution is particularly valuable for platforms operating in emerging markets, where internet bandwidth is limited, and the need for efficient video compression is critical.
  • Global Expansion: The team aims to expand its reach, starting with Africa, where there is a high demand for efficient video processing solutions. Once the model is proven in this market, it will be adapted for other regions globally.

 

The video subnet is positioned to tackle the $400 billion market for video processing, which includes sectors like streaming, cloud storage, and video post-production. The technology will be used in various industries such as surveillance, medical imaging, and industrial video applications. This makes it a versatile solution with the potential to scale across a broad range of markets.

 

MEDIA

A big thank you to Tao Stats for producing these insightful videos in the Novelty Search series. We appreciate the opportunity to dive deep into the groundbreaking work being done by Subnets within Bittensor! Check out some of their other videos HERE.

In this session, the team behind Vidaio provides an in-depth overview of their AI-powered video processing solution. They discuss the challenges facing video content owners and streaming platforms, particularly rising storage and bandwidth costs due to the explosive growth in video content online. The team introduces their core offerings: AI-driven video upscaling to enhance low-resolution videos to HD or 4K and AI-based video compression to reduce file sizes by up to 80% while maintaining perceptual quality. With a focus on solving real-world problems, they explain how the subnet is built on the Bittensor network to allow miners to participate in the processing tasks. The session also highlights the team’s expertise, their roadmap for the subnet’s future developments, and their strategic vision for partnerships and collaborations in the video processing space. The potential impact of this technology is significant, particularly in emerging markets where video delivery and storage solutions are critical.

Novelty Search is great, but for most investors trying to understand Bittensor, the technical depth is a wall, not a bridge. If we’re going to attract investment into this ecosystem then we need more people to understand it! That’s why Siam Kidd and Mark Creaser from DSV Fund have launched Revenue Search, where they ask the simple questions that investors want to know the answers to.

Recorded in July 2025, this episode of Revenue Search features Gareth Howells, founder of Vidaio, the first Bittensor subnet dedicated to AI-powered video compression and upscaling. With 25+ years of experience in video backend systems, Gareth explains how Vidaio is solving major industry pain points: reducing file sizes by up to 80% without perceptible quality loss, and upscaling old low-res footage to crisp 4K using AI.

The compression service—targeting VOD platforms, telecoms, and CDNs—offers massive savings in storage, bandwidth, and battery use, especially crucial for emerging markets. While upscaling has a niche use case (e.g., content archives), Gareth confirms 90% of future revenue will stem from compression. Vidaio plans to charge enterprise clients in fiat via API integrations, with revenue cycling back to support the subnet’s alpha token. Trusted Execution Environments and encryption tech are on the roadmap to ease enterprise security concerns. Compression testing begins within days, with a full market launch expected in under 90 days. Gareth’s vision is clear: deliver better video, cheaper, and redefine how the world experiences content.

Vidaio (Subnet 85) does AI video upscaling (SD → HD/4K) and compression (dramatically smaller files with similar perceived quality). Their consumer web app is live in beta—demo showed ~95% size reduction—with paid tiers coming (think ~$0.05/min, ~75% margins). Beyond creators, the big targets are streamers/broadcasters, legacy libraries, security/medical, and autonomous fleets—anyone drowning in storage/CDN costs. Near-term roadmap adds a streaming pipeline (auto encoding ladders + Hippius storage), plus R&D on colorization and selective video generation/inpainting.

For enterprises needing NDAs and tighter control, Vidaio introduced an Enterprise Track: vetted “Elite miners” execute jobs off-subnet; clients pay fiat split roughly ~75% to miners / ~25% to Vidaio. Miners must post an AlphaBond—locking ~50% of their fiat payout equivalent in ALPHA until client acceptance—creating ALPHA demand/lockups while miners get fiat to cover infra (less sell pressure). Vidaio’s cut first covers OPEX, with surplus flexed between buybacks, product, and growth. Benchmark goals: surpass Topaz Labs (~$8.3M/yr consumer) and Visionular (~$10.5M/yr enterprise) while keeping the subnet as the innovation engine.

NEWS

Announcements

vidaio_τ Follow 266 1,077

vidaio is an open-source, decentralized video processing project, specializing in video up-scaling and compression. we are subnet 85 on the bittensor ecosystem.

vidaio_
Load More

View other subnets

Back to the Directory