How to render Blender without a GPU

If your PC has no GPU and you need a single still image today, your own CPU can still render it; an animation, or anything you want back in minutes, is a cloud job. JANCTION Render renders your .blend, bpy script or 3D file on a datacenter NVIDIA GPU: a 24-frame 720p clip took 59 GPU-seconds (about 6 JPY), Blender's Classroom demo 8.9 GPU-seconds per 1080p frame, measured on the production GPU. A new key starts with a welcome credit of up to 500 JPY (100 JPY with no card, 400 JPY more after registering a card, which is not charged). The other options are compared below by what they cost you in time and setup. AI agents connect at https://render.janction.jp/mcp (remote MCP with OAuth; connecting creates a free key).

Updated 2026-11-12

What it does Your own .blend, bpy script or 3D file (glTF/GLB, FBX, USD, OBJ...) rendered to images or video on a cloud NVIDIA GPU, from an AI app, the terminal or the browser. The first preview needs no payment.
What it does not do It does not speed up the Blender viewport on your machine, does not take non-Blender video, simulations or general GPU jobs, and an AI that is not connected cannot render on its own.

Try your own file for free Build it with an AI and try

Browser only; no connector or CLI needed. A 100 JPY welcome credit for a new key (14 days, no card; one per network every 30 days, further keys on the same network share what is left), and 400 JPY more for saving a card that is not charged (once per card); after that, previews stay free up to 2 GPU-minutes a day and finals cost 0.1 JPY per GPU-second.

Inspect the input and output, then reproduce

For this controlled benchmark, install local Blender 5.2 to prepare the demo and the repository requirements; the script downloads and verifies the public archive. No local GPU is needed. It submits real renders; check pricing and your balance first.

Measured	2026-10-08
Input and license	CC0
Settings	frames 1-3, 1920x1080, 128 samples, denoising on, PNG, Cycles, Blender 5.2
GPU seconds	26.79
Round trip	28.4 s
At the current per-second rate	2.68 JPY
Output	3 PNG frames
Source record and procedure	https://github.com/JasmyLab-JANCTION/janction-render/blob/main/scripts/bench_same_scene.py
Input SHA-256	`0f0ecc58d45f12b2f4272a53e2d8e69135518e16e956b118f10c20d8dcfdf5e6`

python scripts/bench_same_scene.py --scenes classroom --out bench.json

Individual historical observations. Queueing, transfer, input and software versions can change elapsed time. Prices are illustrative, before free credit, whole-yen rounding and outcome ceilings; check /v1/estimate and /v1/pricing for your request.

Decide by what you have

Any GPU of your own that Cycles can use (NVIDIA, Apple silicon, a recent AMD or Intel GPU): render locally. On a simple scene a gaming laptop's GPU matched the production GPU per frame (table below); the cloud only wins when a frame takes longer than you want to wait, the scene does not fit your memory, or you want the laptop free.
One still on a laptop without a usable GPU: the CPU gets there (a recent laptop CPU took 20 seconds for a simple 1080p frame; a heavy scene takes minutes); a datacenter GPU returns the same frame in about 5 GPU-seconds (table below).
An animation: this is where a laptop loses a day. 24 frames at 720p took 59 GPU-seconds here, 100 frames 245.
No budget at all: SheepIt (you render for others and wait in a queue) or a free notebook GPU (you set up Blender yourself each session). Here, the welcome credit covers the first jobs with no card.
An AI assistant that should do the work: it needs a service it can call as a tool. Options that only have a website do not fit that.

The same scenes, measured

Measured on JANCTION Render's production GPU (NVIDIA RTX PRO 6000 Blackwell), 2026-09-30 to 2026-10-01; individual runs, not a guarantee. GPU seconds include Blender start-up and scene loading. The cost column is 0.1 JPY per GPU-second, which applies after the welcome credit. Every row can be reproduced with the scripts on /benchmarks.

scene	frames	size / samples	GPU seconds	round trip	cost
rotating cube (bpy script)	24	1280x720 / 64	59.0	70 s	about 6 JPY
rotating cube	24	1920x1080 / 128	129.4	133 s	about 13 JPY
rotating cube	100	1280x720 / 64	245.4	247 s	about 25 JPY
Poly Haven room (HDRI and models)	24	1920x1080 / 128	136.8	140 s	about 14 JPY
Blender demo "Classroom" (66 MB .blend)	3	1920x1080 / 128	26.8 (8.9 per frame)	—	about 0.9 JPY per frame
4-frame preview of the cube (tiled)	4 of 24	640x360 tiles / 16	4.0	8 s	within the daily free previews
rotating cube on a laptop CPU, no GPU used (Core Ultra 9 386H, 16 threads)	1 (24 estimated)	1280x720 / 64	6.6 s per frame; 24 frames about 158 s	—	no service fee
rotating cube on the same laptop CPU	1 (24 estimated)	1920x1080 / 128	20.0 s per frame; 24 frames about 480 s	—	no service fee
rotating cube on the same laptop's GPU (GeForce RTX 5070 Ti Laptop, CUDA)	1	1280x720 / 64	2.5 s per frame	—	no service fee
rotating cube on the same laptop's GPU	1	1920x1080 / 128	6.0 s per frame	—	no service fee

Per frame, the datacenter GPU took 2.46 GPU-seconds at 720p/64 and 5.39 at 1080p/128 (values from 24-frame batches, including Blender start-up and scene loading; the laptop values are the render of one frame only). So on this simple scene a gaming laptop's GPU is as fast per frame as the production GPU, and the laptop CPU is about 2.7x and 3.7x slower. If you have a usable GPU of your own, rendering locally is the faster choice, and this page says so; the service is for people without one, for keeping the laptop free, for heavy scenes (the production GPU has 96 GB), and for splitting frames across several workers. The laptop rows: ASUS ROG Zephyrus G14 (Intel Core Ultra 9 386H, 31 GB RAM, GeForce RTX 5070 Ti Laptop GPU), Blender 5.2 LTS with factory settings, Cycles with denoising, the same settings as the production worker, measured 2026-10-11; the 24-frame CPU figures are one frame times 24. Classroom is a public file: open it with the settings above and put your own frame time next to 8.9 seconds.

The five options side by side

option	cost	speed	setup	works from an AI agent
CPU rendering on your own machine	no service fee	depends on the scene and the CPU	install Blender	with a local Blender interface
Free notebook GPUs (Colab and similar)	free with limits	GPU access and hardware are not guaranteed	prepare Blender and file transfers in the notebook	agent-assisted notebook work is possible; free managed runtimes restrict remote control
SheepIt (volunteer farm)	free, earn points by rendering for others	depends on the queue	upload through the site	no API
Commercial render farms	paid per GPU-hour or credits; some trial credits	depends on hardware and queue	add-on or web upload	depends on the service's API or MCP support
JANCTION Render	a welcome credit of up to 500 JPY for a new key, then 0.1 JPY per GPU-second	the table above	a connector, pip install, or the browser	yes: MCP tools (`https://render.janction.jp/mcp`) and HTTP API

Colab source (checked 2026-10-02): official FAQ. CPU rendering: Blender manual. API support and rendering speed depend on the service and the workload.

What it costs in practice

Three one-command jobs on a public-domain GLB (the Khronos WaterBottle) and a sample script, run on the production GPU on 2026-10-10. The files they produced are in examples/one-command.

uvx janction-render shots <link to the GLB>: four product shots, 13.9 GPU-seconds, about 1.4 JPY.
uvx janction-render turntable <link to the GLB>: a 360-degree MP4, 85.2 GPU-seconds, about 8.5 JPY.
uvx janction-render render ex12_logo_ring_spin.py --frames 1-48 --engine eevee --output mp4: a 48-frame logo animation, 54.1 GPU-seconds, about 5.4 JPY.

Pricing rules and the welcome credit: /pricing. Limits (frames per job, size, file size, how long results stay): /facts. What the whole trip looks like, from the command to the file: Blender cloud rendering, step by step.

When JANCTION Render is not the answer

A scene that needs more than 3,000 frames in one instruction (send it as several instructions), or that has to stay free beyond what the welcome credit covers (splitting adds no free time).
A .blend saved with a Blender newer than 5.3 (save it from 5.3, or export a bpy script).
Work that needs an availability guarantee: there is none yet (a GPU can be lent to other work; the status page shows how many are available right now).

FAQ

How can a class use Blender on CPU-only school computers?

Use the school PCs for modeling in solid view and send previews or finals to a managed cloud GPU. JANCTION works from /start or an MCP/API connection, so the PCs do not need an NVIDIA GPU; the teacher can cap spend per job and per day. If no paid service is allowed, use CPU rendering or SheepIt instead.

Cycles is very slow on my MacBook. What can I do?

First use the Mac's GPU: Preferences > System > Cycles Render Devices > Metal (tick the GPU), then Render Properties > Device > GPU Compute. Macs with Apple silicon or an AMD GPU can do this; others render Cycles on the CPU. Then cut the work: fewer samples with Denoise on, a lower resolution or a render region for test renders, Persistent Data for animations, and EEVEE when the look allows. If it is still too slow, send the .blend to JANCTION Render and the Mac only uploads the scene: on its NVIDIA RTX PRO 6000, 1920x1080 Cycles stills at 128 samples took 6 to 10 GPU-seconds (0.6 to 1 JPY) and a 24-frame 720p turntable took 58 GPU-seconds (about 6 JPY), measured on the production GPU.

Can I render Blender on a Mac without an NVIDIA GPU?

Locally, Cycles uses Metal on Apple silicon but is slow on older Macs; with JANCTION Render the render runs on an NVIDIA GPU in a datacenter and the Mac only sends the scene.

Is Eevee supported?

Yes. Final renders support engine='eevee'; Cycles is the default. EEVEE shader compilation adds overhead, so Cycles can be faster for small previews. The 48-frame logo animation above was EEVEE: 54.1 GPU-seconds.

Overview · All guides · llms.txt