Import a ComfyUI workflow and run it on a rented GPU from a web form.

Drop the JSON. Missing custom nodes and model files are listed by name before anything runs. The inputs you pick become the form fields and the API body.

Join the list Browse templates

The import report lists each node pack and model file the graph needs.

brand_film_v7.json38 nodes6 of 7 resolved1 missing

Item	Detail	Status
ComfyUI-WanVideoWrapper	commit a41c2e9	Pinned
ComfyUI-VideoHelperSuite	commit f02b7d1	Pinned
comfyui_controlnet_aux	commit 9e1d0aa	Pinned
ComfyUI-KJNodes	commit —	Latest (no commit)
wan2.2_i2v_14B_fp8.safetensors	14.3 GB	Cached
umt5_xxl_fp8.safetensors	6.7 GB	Cached
brand_lora_v3.safetensors	312 MB	Missing

Example import report.

Licence check, example

Wan 2.2 · Apache-2.0FLUX dev · non-commercialbrand_lora_v3 · unknown

A graph that uses a non-commercial model can’t be published as a template until you confirm the licence terms.

The widgets you pick become the fields on the workflow’s form.

graph · 7 nodes

Load Imageproduct_imageCLIP Text EncodepromptLoad Wan ModelWanImageToVideodurationKSamplerVAE DecodeVideo Combine

3 of 7 nodes expose an input. The other nodes stay private.

Brand film · product orbit

product_image

prompt

Slow orbit around the bottle, soft rim light

duration

Run · 12 crsample

A benchmark runs your workflow on 4 GPU types and reports credits per run.

RTX 4090 ran out of memory at 720p. RTX PRO 6000 costs 12 cr per run; H200 takes 52s and costs 19 cr.

GPU	Time per run	Credits per run	Result
RTX 4090	—	—	Out of memory · needs 40 GB or more
L40S	3m 40s	19	Works
RTX PRO 6000	1m 30s	12	Recommended
H200	52s	19	Fastest

Sample numbers for Brand film · product orbit, 3 of 4 GPUs completed. The benchmark price shows on its button before it starts.

Publishing saves a numbered version. Endpoints keep the version they were deployed with.

Edits go to a draft. Restoring an old version copies it into a new draft, so the history stays intact. Moving an endpoint to a new version is a separate step on Deploy.

Brand film · product orbit · versionssample

Draft has changes

v7Switch weights to fp8brand-film endpoint2h ago
v6Expose durationyesterday
v5Add end card3d ago
v4Brand LoRA v31w ago

Questions about importing

Will my custom nodes install?

Node packs from the ComfyUI registry or a public Git repo install at a pinned commit. A pack with no commit is installed at its latest version and marked so you can pin it.

What if a model file is missing?

The import report names it with its size. Upload the file, pick an equivalent, or paste a download link. In the example above, brand_lora_v3.safetensors is missing.

Who can see my graph?

Only you. A shared version can expose the form without the graph.

What does saving or versioning cost?

Nothing. Credits are spent when a workflow runs on a GPU. Rates are on the pricing page.

Which files can I import?

API or UI workflow JSON, or a PNG with the workflow embedded in its metadata.

Send me a ComfyUI graph that won’t import.

Join the list, then reply with the JSON. I’ll send back its import report and the credits per run on each GPU.

Email when signups open · Unsubscribe any time