Real-time AI video from your own ComfyUI graph.

Point a camera at it, change the prompt while it's running, and watch the output change in the same second. On Livepeer. Billed by the second.

Webcam in → live out. No signup.

input

Output appears here

output

Prompt — change it mid-stream

cost $0.000elapsed 0.0sfps 0rate $0.003/s

Local preview demo — cost ticker uses StreamDiffusion retail rate. Production runs land on staging orchestrators and meter via PymtHouse.

Latency is the headline

Warmup is visible. We measure it and own it — time-to-first-frame and steady-state FPS for featured pipelines.

Pipeline	TTFF	Steady FPS	Rate
StreamDiffusion img2img	1.8s	12	$0.003/s
Krea generative	2.4s	8	$0.004/s
Depth control	2.2s	10	$0.003/s
VLM analyze	0.9s	2	$0.002/s

Featured: streamdiffusion-img2img · resolutions are picklist-only (engines baked in image)

1
Import your graph
Drop a ComfyUI workflow JSON built from our supported nodes and models.
2
Run on Livepeer BYOC
A dedicated worker holds your stream (capacity 1). Prompt updates apply mid-flight.
3
Pay billable seconds
Metered from signed network receipts via PymtHouse — to the microdollar.

Same Comfy graph — if every node and checkpoint is in our image. Custom LoRAs and arbitrary node packs are not magic; check the curated list.

Human-verified free tier · $5 included · no agent free credit.