Cloud burst

When local hardware is not enough, the same workflow keeps going.

Pin the models and the look you already trust, match a local reference, then burst on cloud GPUs. You do not have to operate a Comfy graph to get production capacity.

For creators and small teams who outgrew a local GPU and have a real job to ship.

Production GPU fleet. AWS capacity for studio jobs with hard delivery dates

Deadline ops. We have run GPU capacity for studios that cannot slip picture lock

Private pilot environments. Isolated setups scoped to your workflow evaluation

Fixed-scope pilots. From €750. Fee refundable if we cannot reproduce the agreed reference

Built for people who outgrew one GPU

Creators, influencers, and small teams who already ship video and images with open models.

You already have a look or workflow you trust locally

You have a deadline or a recurring production load

VRAM, speed, or machine differences are burning time

A local workstation breaks under production load

The graph that works on one seat gets fragile as soon as you need more capacity or another machine.

Works on one artist’s machine

Can run out of memory during a batch

Custom-node and model versions drift

Capacity is limited to available local cards

Failures may be discovered after hours of compute

Workflow dependencies are recorded

Approved models and custom nodes are isolated

Reference output is validated

Suitable GPUs are benchmarked before the batch

Capacity can extend beyond local hardware

How RenderBob changes the production workflow

Same graph, recorded environment. Dependencies, models, and custom nodes are written down so another box can rebuild the job.

Reference before the live batch. We match a trusted historical output within an agreed tolerance before you run new work.

Capacity with a cost ledger. We benchmark GPUs against your job, then run one non-critical batch with monitoring and a spend record.

What is delivered in the pilot

Dependency manifest. The custom nodes, models, and versions the job needs, written down.

Pinned private environment. An isolated cloud setup matched to that manifest.

Reference-output validation. Proof the pinned environment matches your trusted result within tolerance.

GPU benchmark against the submitted job. Which cards fit, measured before you commit a production batch.

Monitoring for supported runs. Live status while supported jobs run.

Cost ledger for the pilot batch. GPU spend for the supported batch, itemised for finance.

How the four-step pilot works

Submit. Send a completed historical job and the reference output you already trust.

Match. We rebuild and pin the environment, then match the reference within an agreed tolerance.

Batch. Run one non-critical live production batch under real conditions.

Decide. Check match quality, speed, cost, and ops fit. Then choose whether to continue.

Cloud burst

One real workflow, pinned in a private cloud environment, matched to your reference, GPU-benchmarked, and supported for one non-critical production batch.

From €750

Fee refundable if we cannot reproduce the agreed reference workflow.

Common objections, answered

Can you support custom nodes?. Most production graphs use them. In evaluation we review the nodes your workflow needs and tell you what we can pin. Compatibility is never assumed for every custom node.

How are private models handled?. Pilot environments stay scoped to your evaluation. Private models and licensed weights use access controls agreed for the pilot. Retention and transfer steps are confirmed before upload.

Who owns workflows, models, and outputs?. You keep ownership of workflows, models, and outputs. Running a pilot does not give RenderBob rights to your assets.

What happens to uploaded material after the pilot?. [Retention policy pending confirmation] Retention, deletion, or return is agreed up front. Treat this answer as provisional until that policy is approved.

Who is responsible for model licences?. You stay responsible for model licences and usage rights. RenderBob does not grant third-party model licences.

Why does the first batch need to be non-critical?. So you prove the pinned environment before picture lock or a client deadline depends on it.

What if RenderBob cannot reproduce the reference?. If we cannot match the agreed reference within the agreed tolerance, the pilot fee is refundable.

What affects the final price?. Pilots start from €750. Unusual custom-node stacks, large private model sets, tighter match tolerances, or support past one non-critical batch can change the quote. Scope is confirmed before you pay.

How long does evaluation and setup take?. We normally reply within two business days. Setup time depends on workflow complexity and how fast reference materials can be shared securely.

What happens after the pilot?. You review match quality, speed, cost, and ops fit, then decide whether to keep using RenderBob for capacity. Nothing auto-renews.

Which GPUs and regions are supported?. [GPU and region policy pending confirmation] Cards and regions are confirmed during evaluation for your workflow. There is no public matrix on this page yet.

Is this suitable for an individual creator?. We are targeting creators and small teams with a deadline and a look they already trust. Solo creators with that same setup can apply. People looking for free compute should look elsewhere.

Production GPU operations experience

RenderBob already runs production GPU capacity for creative deadlines. Classical 3D work is part of that history. This page is about cloud burst for generative media.

The offer takes one real workflow: pin it, match the reference, benchmark GPUs, then support one non-critical batch when local hardware runs out. ComfyUI stays under the hood.

Ready to pin a look for a real deadline?

No instant GPU rental. Tell us about the job and deadline. We review fit and normally reply within two business days.

Talk about a burst

No instant GPU rental. Tell us about the job and deadline. We review fit and normally reply within two business days.