People-powered AI / BSC

INTELL1GENCE
SH4RED

A shared GPU network for image and video generation. Users create with credits; providers connect local GPUs and earn when renders complete.

Scroll Explore the shared-compute product, worker flow, earning model, and live console. Product below
Project SH4RE / token ticker $f0ur / shared compute on BSC
Product

AI powered by shared compute.

SH4RE should be explained as a real media compute loop, not just a token story. The network starts with a router, GPU worker registry, model capability map, and render receipts.

UserSubmits image, video, upscale, or custom workflow jobs.
RouterMatches job tier to online workers.
WorkerRuns ComfyUI, Wan2.2, SDXL, Flux, or API adapter.
ResultReturns image, video, preview, and progress events.
ReceiptRecords job metadata and rewards.
01 Send a request The user selects text-to-image, image-to-video, Wan2.2 video, upscale, or a custom ComfyUI graph from the SH4RE app. Image / video / workflow
02 Route by capability The router checks online workers, current load, model profile, and measured speed before selecting a worker. Registry + queue + selection
03 Stream the result Progress, queue state, preview frame, and final media URL stream back through the router. Progress-first UX
04 Reward useful work Workers earn credits based on verified completion, speed, uptime, and model tier. Later settlement can use $f0ur. Useful compute first
Router

The router is the network core.

The first serious version needs one reliable middle layer: it authenticates users and workers, keeps the render queue, tracks online GPU capacity, selects a worker, and returns media results.

Router responsibilities
Wallet/session identity for users and workers
Incoming image/video jobs, selected tier, timeout, retry state
Online GPUs, model support, VRAM, speed, load, uptime
Weighted selection by measured throughput and availability
Live network status: workers online, jobs routed, videos rendered
Important: prompts do not need permanent storage. The router can move jobs, return results, and keep only receipts.
Requester experience
Pick task type: image, image-to-video, video, upscale, or custom workflow.
See estimated GPU tier, credit cost, and expected latency.
Watch progress events, preview frames, and final media return through the app.
Receive a receipt with worker tier, route, and completion metadata.
Workers

Turn local GPUs into media workers.

SH4RE starts with image and video generation. A worker publishes GPU model, VRAM, supported pipelines, uptime, and accepted price. The router sends jobs only when that machine is online.

GPU classSuggested workloadsExample modelsReference payout
RTX 3060 / 4060 12GBSDXL image, LoRA, product shots, low-res previewsSDXL, Flux schnell, Qwen image lite$0.015 - $0.05 / image
RTX 4070 Ti / 4080 16GBHigh-res image, batch images, short image-to-video queueSDXL, Flux dev, Wan2.2 480p jobs$0.04 - $0.16 / render
RTX 4090 24GBWan2.2 video, HunyuanVideo, larger ComfyUI graphsWan2.2 T2V/I2V, CogVideoX, upscale$0.25 - $1.20 / video
RTX 5090 / A6000 32GB+Longer clips, multi-stage workflows, premium low-latency queueWan2.2 720p, heavy ControlNet, video refine$0.80 - $3.60 / job
Pricing should follow public API market pricing as a ceiling, then split user payment into provider payout, platform reserve, retries, bandwidth, and token incentives. The table is a launch reference, not guaranteed income.
Create Cloud

Connect once, earn when jobs finish.

In production, a provider runs a small SH4RE worker client next to ComfyUI or another media runtime. The client keeps an outbound WebSocket open, receives signed jobs, calls the local model, and uploads the result back to the router.

01 Install runtime Install NVIDIA driver, CUDA, Python, ComfyUI, and the required model files such as Wan2.2, SDXL, Flux, or LoRA packs. Local GPU stays local
02 Run worker client The SH4RE client benchmarks the GPU, reads VRAM, checks installed model names, and registers capability metadata. Outbound connection only
03 Accept routed jobs The router sends a prompt, template, seed, resolution, and timeout. The worker calls local ComfyUI or your API adapter. No public port required
04 Return receipt Completed jobs return result URL, duration, GPU tier, payout estimate, and uptime score. Failed jobs can retry elsewhere. Auditable completion
WorkerGPUModelsStatusToday
pink-node-4090RTX 4090 24GBWan2.2 / SDXL / upscaleonline$18.42 / 9h
four-lab-4080RTX 4080 16GBSDXL / Flux / I2V previewonline$7.90 / 5h
wan-rig-a6000RTX A6000 48GBWan2.2 720p / video refineonline$31.70 / 11h
home-3060RTX 3060 12GBSDXL / LoRA imageoffline$2.16 / 1h
Token Layer

$f0ur supports the compute market.

The token should not be the product. The product is shared compute. $f0ur can become the coordination layer for worker incentives, settlement boosts, launch liquidity, and public receipts once the network proves demand.

Access Users can hold credits for requests first; token access can come later through discounts, staking, or priority routing. Keep access simple
Incentives Workers can earn credits now and later receive $f0ur boosts based on verified completion, uptime, and score. Useful work first
Launch Flap can be used to launch $f0ur, but the website should keep routing users through the SH4RE product story. Launch after product framing
Receipts After deployment, this page should link to the token contract, BscScan, pair, treasury, and live reward routes. Visible addresses
Launch App

SH4RE app console

The operational console lives on its own page. Use it to submit image/video jobs, configure the upstream generation API, simulate routed GPU workers, and preview provider payouts.

01 ComfyUI Cloud Submit prompt, model, size, and workflow type. In the early build, SH4RE can call one backend API while displaying routed workers. Image / video MVP
02 Worker Profile Simulate GPU capability, online state, uptime, supported models, and provider payout before distributed clients ship. Contributor onboarding
03 Router Demo Simulate queue, worker selection, streaming output, and local receipts before backend deployment. Network preview