Skip to main content
TsugiCinema Inc. · Delaware C-Corp.

TsugiCinema

“home theater rewoven”

The highest-fidelity place on the internet to watch and publish AI-generated cinema. Live today on a $200 device that drives any modern 4K TV past what streaming and Blu-ray can carry.

Join the Waitlist See how it works
Patent foundation
Three filed US provisional patents
8.0 to 9.2 / 10 external patentability scoring
On consumer silicon
12-bit 4:4:4 pipeline, live today
Off-the-shelf $200 device · ASIC reference design ready
Strategic posture
Open to silicon, OEM, and distribution
AOMedia AV1 / AV2 · UALink Consortium · NVIDIA Inception
Demo available
San Francisco reference room
Sony Bravia 8 II · KEF acoustic suite · by appointment
The problem

Your TV is underserved.
We have already solved it.

Consumer displays have outrun the software that drives them. Modern 4K TVs handle smoother gradients, deeper color, and cleaner motion than the platforms ever ship to them. The services you watch on still send 8-bit 4:2:0 at commodity bitrates because that is what YouTube was calibrated for, more than a decade ago.

AI-generated video has different statistical properties from camera-captured footage. That compression destroys the subtle gradients, the temporal coherence, and the fine textural detail that defines whether an AI short looks like cinema or like slop.

We ship master-quality 4K to consumer TVs through a $200 device today, in a real customer's home. The result is visibly better on any modern 4K TV, and reference-grade on premium panels. You can see it in person in San Francisco, or in your own home once you are in the alpha cohort.

Side-by-side: the same sunset scene, master quality on the left with smooth gradient and sharp red title, typical streaming on the right with visible color banding in the sky and chroma bleeding around the title.
The same scene, two ways. Smooth color and sharp edges on the left are what your modern 4K TV can resolve. The visible banding and color bleeding on the right are what most streaming actually sends.
What we built

Three patent-pending pillars,
One Product.

The three filed US provisional patents below sit alongside each other as a portfolio bundle. Two are running on consumer silicon today; the third is the substrate for the long bet.

Diagram: how three streams (carrier, color delta, gray delta) combine into one master-quality picture

Dual-layer compression

Master-quality 4K 12-bit 4:4:4 from a 30 to 38 GB file, approximately 40 percent smaller than AV1 at the same perceptual quality. Running on consumer silicon today.

Patent-pending · US Prov. 64/054,446
Trinity decoder system block diagram with three slave clock domains feeding an FPGA digital logic device

Plesiochronous synchronization

A multi-decoder synchronization fabric that joins two or three decoders, scanline aligned, to deliver 12-bit color from commodity 8-bit-per-decoder silicon. The Kintsugi seam, in hardware.

Patent-pending · US Prov. 63/987,139
Infinity gradient-consensus supercomputer system block diagram with master clock domain, PCIe Gen 5/6 + CXL 3.1 switch fabric, and shared memory pool

Distributed AI training substrate

A sub-nanosecond gradient-consensus architecture for distributed AI training. Apparatus claims filed; software-analog validation across vision and large language model continual learning.

Patent-pending · US Prov. 64/055,093

Specific claim language is counsel-protected. Strategic partners under NDA receive the full technical annex on request via partnerships@tsugicinema.com.

Technology

Three filed US provisional patents.
One working consumer device.

A vertically integrated stack from silicon to network to AI substrate. The patent portfolio below is independently externally validated. Specific claim language is counsel-protected; strategic partners under NDA receive the technical annex on request.

Filed provisionals
0
Duality, Trinity, Infinity
Reference rig
$0
Off-the-shelf consumer device
Bit depth shipped
0-bit
4:4:4 chroma, 4K
Creator split
0%
to creator, every dollar

Patent portfolio

64/054,446
Duality
Keyed video compression with a cryptographically separated companion stream
Filed 2026-05-01
63/987,139
Trinity
Plesiochronous multi-decoder video synchronization
Filed 2026-02-19
64/055,093
Infinity
Sub-nanosecond plesiochronous gradient consensus for distributed AI training
Filed 2026-05-01
"Exceptionally well-positioned. Robustly novel. A non-obvious leap over the established visual-cryptography teaching."
Duality · Deep Research evaluation
"Structural ingenuity in the Sideband Flow Control Mechanism. An examiner would struggle to find a motivation to combine prior art."
Trinity · Deep Research evaluation
"The strongest patentability posture in the portfolio. A profound architectural divergence from prior cross-domain references."
Infinity · Deep Research evaluation

Patent landscape and freedom to operate

Demos and Demografx published the foundational resolution-layering patents in the 1990s. Those expire in 2026, opening a clean freedom-to-operate window. Trinity uses additive-only reconstruction, fundamentally different from Dolby Vision's predictive approach.

Patent landscape and freedom-to-operate timeline. Demos US 5,661,524A (resolution layering, filed 1996) and Demografx prior art protect 1996 through 2014. Dolby Vision launches in 2014 using predictive reconstruction. AV1 codec ratifies in 2018 with $0.20 to $0.26 per-unit royalties, financially immaterial at $899 to $999 retail. Demos US 5,661,524A expires in 2018. The remaining Demos and Demografx patents (US 10,297,008B2, US 8,737,466B2) expire in 2026, opening the freedom-to-operate window for Trinity. Trinity provisional patent is filed at the start of that window with non-provisional plus Track 1 prosecution projected for 2026 through 2028.

The stack, four layers

Layer 01

Silicon

Off-the-shelf $200-class consumer SoC. Two-context decoder pipeline driven by a custom dual-layer compression scheme. Trinity FPGA reference for 3-decoder scale.

Stylized top-down view of an off-the-shelf consumer SoC with a two-context decoder pipeline feeding a dual-layer compositor block.
Layer 02

Pipeline

Master-quality 4K 12-bit 4:4:4 ingest. Custom keyed compression with a cryptographically separated companion stream. 30 to 38 GB per 90-minute master. Approximately 40 percent smaller than AV1 at parity perceptual quality.

Duality keyed compression pipeline: a master 4K file is split into a visible stream and a cryptographically keyed stream, then recombined at decode into a 30 to 38 GB output approximately 40 percent smaller than AV1.
Layer 03

Network

Residential edge nodes function as delivery peers, storage nodes, and (Phase 2) compute nodes. Hybrid origin and edge fabric. No crypto-speculation cold start.

Residential edge mesh: a central origin node connected to six residential edge nodes by gold seams, with dashed edge-to-edge peering lines.
Layer 04

AI substrate

Sub-nanosecond plesiochronous gradient consensus apparatus for distributed neural network training. Apparatus-claim provisional filed; software-analog continual-learning validation across vision and large language model workloads.

Infinity AI substrate: four compute nodes ringed around a central consensus device, joined by a solid White Rabbit timing backbone and a dashed out-of-band phase-correction sideband.
Live, not theoretical

A working consumer device, on a $200 box, today.

The dual-layer compression pipeline runs on off-the-shelf consumer silicon, driving a reference Sony Bravia 8 II in our San Francisco demo room. We can show you in person. Email partnerships@tsugicinema.com.

  • ·San Francisco reference room (Sony Bravia 8 II)
  • ·Loaner TsugiNode for qualified pilot partners
  • ·Technical walkthrough for strategic partners (post-NDA)
San Francisco reference room: Sony Bravia 8 II reference TV between two KEF floorstanding speakers, playing Blade Runner 2049 at night. A sunset projector casts a warm halo on the wall above a curved bouclé sofa; floor-to-ceiling industrial windows look out on the SF skyline with Salesforce Tower lit blue in the distance.
San Francisco reference room
Partnerships

We are open to silicon, OEM, distribution, and studio conversations.

The TsugiNode reference design is consumer SoC today and ASIC tomorrow. The Duality pipeline is ready for SoC manufacturer integration. The Infinity gradient-consensus substrate is targeting the UALink Consortium track. If your roadmap intersects any of those, the door is open.

Silicon and OEM partnership icon: a SoC die at center bonded out to a surrounding integration frame.
Silicon & OEM

SoC integration, NPU integration

Snapdragon-class NPUs, AV1 / AV2 ecosystem, dual-decoder SoCs, GPU-shader sync. Reference-design walkthroughs and technical-annex packages available under NDA.

Distribution partnership icon: a stylized home cinema room with a wall display flanked by two stand speakers.
Distribution

Architect, installer, retail

Premium home-cinema install channel. Cofounder Dylan brings architect-channel relationships from Apple flagship retail and Edge at Hudson Yards. We are quoting full rooms today.

Studios and research partnership icon: a film slate joined by a gold seam to a research bar-and-trend graph.
Studios & research

AI studios, B2B compute

Selectively curated AI film studios for the platform. B2B compute for research workloads on the residential edge. First named B2B prospect: Harvard PhD computational biology.

Strategic conversations

Strategic-partner inbound is read directly by the founder. Technical annex, patent-portfolio detail, and reference-room demo offer follow under NDA.

partnerships@tsugicinema.com
Who we serve

For the picture.
For the maker.

For viewers

Your TV deserves more.
Your box delivers it.

Our dual-layer pipeline ships exactly what your hardware can show. Cleaner color, smoother gradients, and lower compression noise are visible on any modern 4K TV. The headline number (12-bit 4:4:4) is unlocked on reference-grade panels. The advantage disappears once it gets re-compressed by YouTube, so the honest test is on the box itself.

Viewer value proposition: pay direct to the creator who made it; never see a forced ad you did not choose; your box earns credits for premium content while it sits idle.
Read how playback works →
For creators

Your master, shipped at master quality.
Ninety cents of every dollar.

You made something that looks like what you rendered. Why does the rest of the internet make it look like something else? TsugiCinema ships your master at master quality. You set the price. Your superfans know exactly what they are paying for.

Creator revenue split donut: 90 percent to creator (gold), 10 percent to platform (charcoal).
Forced ads
0
ever, on any video

Human creators are held in the highest esteem. AI studios are selectively curated to a published quality bar.

For human creators → For AI studios →
Brand commitments

Three commitments.
No exceptions.

Above the Fidelity Floor.

Every video on TsugiCinema clears a published quality threshold. Specific, measurable, documented. No other platform commits to that.

Never a forced video ad.

Pre-roll, mid-roll, post-roll, interstitial: all forbidden. Forever. Display ads on browse pages only, clearly labeled, never on the page where you are watching.

Pay only for what you watch.

No platform subscription. No paid tier to remove ads (there are none to remove). Creators set their own prices. You know exactly what you are paying for.

"When viewers think a creator's price is too high, the outrage is at the creator, not at the platform. That is the correct architecture."

The long bet

We put the power of compute, storage, and AI
back in the hands of the people.

Residential edge mesh: a central origin node connected to six residential edge nodes by gold seams, with dashed edge-to-edge peering lines.

Every TsugiNode in a home is also a delivery peer, a storage node, and eventually a compute node. Operators choose which artists, studios, and research customers receive their spare compute. The platform does not capture that decision.

Large private AI is consolidating compute into a small number of data centers. There is an alternative architecture that scales residential, on the margin that matters: power, cooling, latency to the user, and consumer agency.

This is the scalable answer to data-center-bound AI infrastructure, and it is the reason the same hardware that drives your TV can also fund an open, consumer-owned compute grid.

Read the full vision →
Phase 1

Delivery peer

Edge CDN. Reduces origin egress and improves last-mile latency.

Phase 1

Storage node

Operator-controlled cache. Gifted unlocks, master-quality preloads.

Phase 2

Compute node

Spare cycles for AI inference. Operator chooses which jobs run.

Phase 2

B2B compute

Research, studio, biology workloads. First named prospect: Harvard PhD computational biology.

The team

Two cofounders, one fractional CoS.
All equity-only at pre-seed.

NVIDIA Inception Program Inception Member
Tong Liu

Tong Liu

Cofounder & CEO

Hardware architecture, FPGA and ROM engineering, dual-layer compression. Sole inventor on the three filed US provisional patents. Former Google Search accessibility lead.

Dylan Lindeberg

Dylan Lindeberg

Cofounder & COO

Operations, GTM via the home-cinema install channel, audio engineering. Architect-channel relationships through prior work on Apple flagship retail and Edge at Hudson Yards.

Wenling Ma

Wenling Ma

Fractional Chief of Staff & Head of Compliance

Three decades of governance, strategic planning, and copyright clearance experience. Money-transmitter regulatory analysis and Phase 3 studio licensing governance.

Strategic-partner conversations and named technical advisors disclosed under NDA. See about for the company story and the Kintsugi origin.

Waitlist

Join the waitlist.

Tell us who you are. We will be in touch.

  • ·Alpha tester / consumer
  • ·Home theater bundle (premium install)
  • ·Human creator (filmmakers, EDM artists, individual authors)
  • ·AI film studio (selective)
How to join

Send a short note to the address that fits you best. We read every email.

Strategic-partner inbound is read directly by the founder. We respond within a few business days.