CloudsForgeCloudsForge
CloudsForge

Want to generate videos
with Open-Source models but don't have the GPU or the ability to set them up?

CloudsForge is a desktop app that rents you a cloud GPU and configures everything in a couple of clicks. You run state-of-the-art open-source video models at the GPU provider's rate

You own what you buyMulti-cloud, multi-modelZero telemetry
The Problem
Without CloudsForge

Running Open-Source AI is unnecessarily hard.

Whether it's video, image, or audio generation, you're stuck configuring conda environments, resolving CUDA mismatches, chasing dependency conflicts, deciphering multi-page install guides, and figuring out which cloud GPU providers to use and at what price.

condaCUDA 12.1pip conflictsPyTorchCUDA drivers40+ dependencies
bash — opensora-install
$conda create -n opensora python=3.10
$conda activate opensora
$pip install torch==2.1.0+cu121 --extra-index-url ...
Collecting torch==2.1.0+cu121
Downloading torch-2.1.0+cu121... [━━━━━━━━━━ ] 47%
$pip install -r requirements.txt
ERROR: Dependency conflict: transformers 4.36.2 requires ...
$pip install transformers==4.35.0 --force-reinstall
... 23 more packages ...
$python -m opensora.train --config configs/opensora/train/64x512x512.py
CUDA error: no kernel image is available for execution
$
There's a better way
Cloud setup✓ Done
Install CloudsForge
Add your cloud API keys
Rent a GPU & let us set it up
Generate video
With CloudsForge

Four steps. No terminal required.

CloudsForge connects to cloud providers and handles all the remote dependency management, environment setup, and GPU configuration automatically. You add your API keys. You rent a GPU. You generate video. That's it.

Features

Only the functionality that you actually need.

CloudsForge is purpose-built for running open-source video models on rented cloud GPUs without the setup hell.

One-Click Install

No conda, no SSH, no terminal. CloudsForge automatically sets up the full environment on your rented GPU packages, dependencies, model weights.

Model Manager

Browse and switch between Open-Sora, Wan, HunyuanVideo, and more. With new models added as they're ready. Open-Sora v2 and Wan2.1 and HunyuanVideo available today.

Job Queue

Queue multiple generation jobs across cloud instances. Track progress, stream logs in real time, cancel jobs, review history.

GPU Provider Browser

Browse GPU offers from RunPod and Vast.ai filtered by model requirements. Compare real prices. Create and manage instances without leaving the app.

Clean Interface

Prompt, settings, preview, queue and everything is exactly where you expect it. Designed for repeated daily use, not just demos.

Zero Markups

You see the provider's real price. You pay them directly. We never take a cut, referral fee, or markup. Our job is to connect you, not profit from the connection.

Roadmap

One workstation. Many surfaces.

Video first. Image, audio, and fine-tuning next. Added to the same app, included with your current major version. These are intentions, not commitments.

Available nowshipping
  • Video generation (Open-Sora v2, Wan2.1, HunyuanVideo)
  • RunPod + Vast.ai integration
  • Job queue, real-time logs, gallery
  • Spending and Benchmarks (beta)
Next 6 monthsin progress
  • Additional video models (Mochi, CogVideoX)
  • Additional cloud providers
  • Studio tier — general availability
Beyondplanned
  • Image generation models (SDXL, Flux, others)
  • LoRA fine-tuning UI
  • More providers, more models

Personal license holders receive everything added within their major version at no additional cost.

Discord

Bugs, ideas, and your generations.

Drop in — chat with other creators, get early access to features, and shape what ships next.

Join the server
How It Works

Up and running in minutes.

01

Install CloudsForge

Download the app for your platform. Run it once.

Windows · macOS · Linux
02

Add your cloud API keys

Enter your RunPod or Vast.ai API key. CloudsForge shows you available GPUs at their real published prices — you compare and choose.

RunPod · Vast.ai
03

Rent a GPU & let us set it up

Pick a GPU offer. One click starts the instance and automatically installs everything: conda environment, dependencies, and model weights (Open-Sora v2 or Wan2.1). Real-time progress shown in the app.

Automated setup · No terminal
04

Generate

Write your prompt or upload a reference. Hit Generate. Your rented GPU does the work. Download your output. Terminate the instance when done.

Runs on your rented GPU. You pay the provider directly.
$5
Free credit
New to RunPod?

Sign up through our link and start for free

$5 is enough to spin up a GPU instance and generate a few test videos at zero cost — try CloudsForge end-to-end before you commit.

Stop Renting Software That Rents You Hardware

Rent GPU at cost.

Cloud AI tools like Runway, Kling, Midjourney, and DALL·E buy GPU time in bulk and resell it to you — marked up, credit-gated, and subscription-locked. CloudsForge connects directly to providers like RunPod and Vast.ai so you pay their actual rate and keep full control across video, image, and beyond.

Pay per hour

Unlimited generations

No credits, no generation limits. You pay for GPU time by the hour at provider rates. Run 50 jobs or 5 — the cost is transparent and direct.

No markups

Real GPU prices

RunPod and Vast.ai publish their rates. We show you those rates unmodified. We don't take a commission, referral fee, or markup. Our job is to connect you, not profit from the connection.

Lifetime access

One-time $59

Pay once, use forever. No monthly fees, no seat licenses, no usage tiers. The license is yours permanently.

Zero telemetry

No data collection

Your prompts, your videos, your ideas — none of it reaches us. No telemetry, no analytics, no usage tracking.

Parallel jobs

Multi-instance parallelism

Run jobs across multiple cloud instances simultaneously. The job queue handles routing automatically.

$0 for models

Free open-source models

Open-Sora, Wan, Mochi, HunyuanVideo — all open-source, downloaded directly from HuggingFace. You never pay for the models themselves.

Comparison
Cloud Tools
CloudsForge
Monthly cost
$30–150/mo
$0 after one-time $59
Generations
Limited by credits
Unlimited (pay GPU time only)
Privacy
Uploaded to servers
Never sent to us. Zero telemetry.
Queue wait
30s–10min (platform queue)
Your GPU, your queue
Offline use
Internet required
Internet required (GPU is remote)
Customization
Platform presets
Full control
GPU cost
Marked up in credits
Native provider pricing
Model cost
Bundled into subscription
Free (open-source)
Model coverage
One platform's models
Any open-source model on the workstation
What We Will Never Do

What CloudsForge Will Never Do

1

Charge you more for GPU access.

We show you available providers and their real prices — you compare, you choose, you pay them directly. We never touch that transaction.

2

Charge you for software you already own.

The version of CloudsForge you purchased is yours, at the price you paid, forever. Future major versions are separate, optional upgrades — never required, never automatic, and your existing version keeps working indefinitely. Optional add-ons exist (like Cloud Queue), but they're never required to use what you bought.

3

Convert your license into a subscription.

CloudsForge Personal is a one-time purchase, and that will not change. We will never retroactively shift purchased licenses to recurring billing or paywall features that were available in the version you bought.

4

Watermark or interfere with your output.

No logos, no branding, no hidden modifications. What your GPU generates is exactly what you get — unaltered, yours completely.

5

Use coins, tokens, or credits.

All pricing is straightforward USD. No artificial currencies designed to obscure what you're actually spending.

vs. cloud tools

Stop paying the subscription tax.

Runway, Kling, Pika, and Sora all rent you access to hardware you don't own and charge per generation. CloudsForge flips that model entirely.

starting at ~$1.39 / hr

Rent GPU at native prices

Use your own hardware, or rent from providers like Vast.ai or RunPod at rates like $0.09/hr. CloudsForge connects directly. We don't touch the transaction, mark up the cost, or take a cut.

$0

Open-source models, always free

HunyuanVideo, Wan T2V, Mochi, Open-Sora — all open-source, all free, downloaded straight from HuggingFace. One click in the model manager. No licence fees, no paywalls, no surprises.

$59 once

One payment. No further charges.

$59 gets you the application. That's the only money we ever ask for on Personal — no seat licences, no renewal, no "pro tier". The licence is yours permanently.

Pricing

Simple pricing. No surprises.

Pay once. Own it forever. No subscription, no markups, no telemetry.

One-Time · Personal

Personal

Pay once. Own it forever.

$59one-time
  • Desktop app for Windows, macOS, Linux
  • Multi-cloud GPU integration (RunPod, Vast.ai, more coming)
  • All current open-source video models (Open-Sora v2, Wan2.1, more)
  • Image, audio, and fine-tuning support as added (within current major version)
  • Job queue with real-time log streaming
  • Video gallery with download
  • Lifetime access to current major version
  • Free updates within current major version
  • Future major versions: optional paid upgrade (~$29 for existing license holders)
  • No subscription, ever
Why paid major versions?

Maintaining CloudsForge across a moving open-source ecosystem is real ongoing work. Major version upgrades fund that work without subscriptions or markups. Your current version is always yours.

Models are not included. They are downloaded separately from official sources (HuggingFace, GitHub).

💬Request features directly
♾️Lifetime Personal license
FAQ

Common questions.

Everything you need to know before purchasing. More in the docs.

No. Open-Sora is an open-source AI video generation model developed by HPC-AI Tech. CloudsForge is a completely separate commercial product — a graphical desktop application that makes it easier to install and run Open-Sora and other open-source models on rented cloud GPUs. We are not affiliated with HPC-AI Tech.

You don't need a GPU in your own machine. CloudsForge connects to RunPod and Vast.ai so you rent a GPU in the cloud. The models require 48–80 GB VRAM — only available via cloud providers at this time. You pay the provider directly by the hour.

The actual AI models (Open-Sora v2, Wan2.1 T2V-14B, and others as they're added) are open-source and free, downloaded directly from HuggingFace through the app.

No. CloudsForge Personal is a one-time purchase. You pay once and own the license permanently. There are no monthly fees, no usage limits, no seat charges, and no renewal required. Updates within the current major version are free. (Studio is a separate, optional annual product for teams — Personal pricing is independent of Studio pricing.)

Yes — for major version releases (roughly every 12–18 months). Your existing version is yours forever and keeps working. Upgrading is optional, and existing license holders get upgrade pricing (typically ~50% off new-purchase price).

Studio is built for teams who need shared prompt libraries, SSO, multi-user job queues, audit logs, and a support SLA. It's a different product for a different buyer — Personal users don't need it, and Personal pricing is independent of Studio pricing.

Yes. Image generation, audio, and LoRA fine-tuning are on the roadmap. Where feasible, these are added within the current major version — meaning Personal license holders get them at no extra cost.

An optional add-on for users who want generations to run while they're away from their machine. The core CloudsForge app works fully without it. Even with it, your prompts and outputs are not stored or analyzed by us — we only manage the job lifecycle on your behalf.

The app requires an internet connection to communicate with RunPod and Vast.ai APIs and to execute jobs on remote GPUs. Your prompts and video outputs are never sent to us — but they do pass through your chosen cloud GPU provider under their terms.

Yes. All models supported by CloudsForge are open-source projects released under permissive licenses (Apache 2.0, etc.) that explicitly allow personal and commercial use. CloudsForge itself is a commercial software wrapper. Always check the specific license of each model for your use case, particularly for commercial output.

CloudsForge runs on Windows 10/11, macOS, and Linux. Because video generation runs on rented cloud GPUs (not your local machine), you don't need an NVIDIA GPU in your own computer.

We provide documentation, a GitHub issues tracker for bug reports, and direct email support for license holders. We actively maintain the software and release updates to keep pace with new model versions and upstream changes.

Yes — that's the primary way CloudsForge works. You connect your RunPod or Vast.ai account, browse available GPU offers (filtered by what each model needs), and rent one directly at the provider's published rate. CloudsForge handles setup and job execution. You pay the provider directly — we're not involved in that transaction.

No. There is no telemetry in CloudsForge. We do not track generation counts, prompt content, video output, model usage, or anything else. Once the app is running, it operates entirely on your machine. We have no server that receives data from your installation. This is not a policy that might change — the app is architecturally incapable of sending this data because we never built the infrastructure to receive it.

Open-Sora v2 and Wan2.1 T2V-14B are available today. Both support Text-to-Video and Image-to-Video. Mochi, HunyuanVideo, CogVideoX, and AnimateDiff are in the pipeline and will be added as support is ready.

Still have questions?

Check the documentation or reach out directly.