The Civitai alternative
Run any LoRA. Train your own.
Bring your LoRAs to FLUX, Qwen Image, Z-Image, Wan, LTX and MiniMax H3, train new ones from a folder of images, and generate images and video with the newest models. Fast results, pay as you go, and an API for all of it.
Your LoRAs, running fast.
Fast.
Blazing-fast inference on every model. Your images and videos come back fast, in the studio and through the API.
Affordable.
Images from $0.024, videos from $0.05 per run. Pay only for what you generate, with no subscription and no monthly fee.
- LoRA-ready models
- 66
- LoRA trainers
- 13
- AI models
- 1,062+
- Monthly fee
- $0
Run LoRAs on today's base models
LoRA endpoints for FLUX, Qwen Image, Z-Image, Wan, LTX and MiniMax H3. Paste the URL of a LoRA file, set its strength and generate; many endpoints let you stack more than one.
Civitai: A large community library of checkpoints and LoRAs, with an on-site generator for running them.
Train image and video LoRAs
Upload a zip of images, add a trigger word and get a LoRA file you can use right away on the matching LoRA endpoint, for FLUX, Qwen Image, Z-Image, Wan and LTX.
Civitai: An on-site LoRA trainer for SD 1.5, SDXL, Flux and some video models, paid in Buzz.
Frontier image and video in one place
Seedream, Nano Banana, GPT Image, Seedance, Wan, Kling and MiniMax H3 sit right next to your LoRAs, on the same balance.
Civitai: Image and video generation built around community models, plus some third-party models.
An API for every model and trainer
Every LoRA endpoint, trainer and generator is one REST call with a single key, at the price shown on its model page.
Civitai: A public API for browsing its catalog; generation runs through a separate developer API metered in Buzz.
LoRA models
Pick a base model, add your LoRA and generate images, edits or video. 66 models support LoRAs today.
![FLUX.1 [dev] LoRA sample output](https://static.wavespeed.ai/media/images/20260408110125_qaxxdycm.webp)
FLUX.1 [dev] LoRA
ImageFrom $0.015

Qwen Image 2512 LoRA
ImageFrom $0.025

Z-Image Turbo LoRA
ImageFrom $0.01

FLUX.2 Klein 9B LoRA
ImageFrom $0.015

Qwen Image Edit 2511 LoRA
Image editFrom $0.025

Wan 2.2 Image-to-Video LoRA
VideoFrom $0.35

LTX-2.3 Image-to-Video LoRA
VideoFrom $0.10

MiniMax H3 Text-to-Video LoRA
VideoFrom $0.125
Train your own LoRA
A character, a product, an art style or a motion: teach it to a model in one run. 13 trainers are ready for image and video.
- 1
Upload your images
Zip a set of images, with optional caption files, and pick a trigger word.
- 2
Train
Choose a trainer and the number of steps. The price is shown before you start and scales with the steps.
- 3
Generate with it
Use the finished LoRA on the matching LoRA endpoint, in the browser or by API.
Image and video, on the newest models
When you don't need a LoRA, the leading image and video models are one click away.
Nails the prompt, and the text in it
Edit and compose any image by asking
Crisp, high-res photoreal images
Fast photoreal images at a low price
Cinematic shots with sound, from a prompt or a still
Expressive characters, quick turnaround
Animate references, edit footage with words
Fluid motion and multi-shot stories
Made on WaveSpeedAI
Every image and clip here came out of models on WaveSpeedAI.
/filters:quality(82)/examples/323fc3db241d9b56f46c445f2ab334c0/1783684221888696489_duh4aXLA.webp)
/filters:quality(82)/examples/fe87ceb035539d65876550870726b9ae/1783684223864317780_aPFXawP6.webp)
/filters:quality(82)/examples/74a13d9ab631f600a9c6633fe3768e91/4bfd0e16-1100-4c3c-b5f4-68db8258aef0-40766a57bec6.webp)
/filters:quality(82)/examples/909134b98b5bd85163a0936ad0da0cdb/1783683528516769834_gGPTSUVN.webp)
/filters:quality(82)/examples/0cbef4bfd187825951ff9cb2c0787280/1783684219348367405_p8cYLApe.webp)
/filters:quality(82)/examples/440dff915c96090b6178ac2046980e86/53bddec8-5e91-4eec-b460-735089911eb5-b42d2013c87f.webp)
Build it into your app
Everything on this page is also an API. Train a LoRA, then call it from your app, script or agent with the same key.
- One REST call per model, with Python and JavaScript SDKs
- Pass LoRAs in the request as a list of file URLs and strengths
- A CLI and an MCP server to hand the same models to your agents
POSThttps://api.wavespeed.ai/api/v3/wavespeed-ai/flux-dev-lora
curl -X POST "https://api.wavespeed.ai/api/v3/wavespeed-ai/flux-dev-lora" \
-H "Authorization: Bearer $WAVESPEED_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"prompt": "A cozy reading nook in watercolor style",
"loras": [
{
"path": "https://example.com/your-style-lora.safetensors",
"scale": 1
}
]
}'
# → {"data": {"id": "<prediction id>", ...}}
curl "https://api.wavespeed.ai/api/v3/predictions/<prediction id>/result" \
-H "Authorization: Bearer $WAVESPEED_API_KEY"import os, time, requests
headers = {"Authorization": f"Bearer {os.environ['WAVESPEED_API_KEY']}"}
task = requests.post("https://api.wavespeed.ai/api/v3/wavespeed-ai/flux-dev-lora", headers=headers, json={
"prompt": "A cozy reading nook in watercolor style",
"loras": [
{
"path": "https://example.com/your-style-lora.safetensors",
"scale": 1
}
]
}).json()["data"]
while True:
result = requests.get(f"https://api.wavespeed.ai/api/v3/predictions/{task['id']}/result", headers=headers).json()["data"]
if result["status"] == "completed":
print(result["outputs"])
break
if result["status"] == "failed":
raise RuntimeError(result.get("error"))
time.sleep(2)const headers = { Authorization: `Bearer ${process.env.WAVESPEED_API_KEY}`, 'Content-Type': 'application/json' }
const { data: task } = await fetch('https://api.wavespeed.ai/api/v3/wavespeed-ai/flux-dev-lora', {
method: 'POST',
headers,
body: JSON.stringify({
"prompt": "A cozy reading nook in watercolor style",
"loras": [
{
"path": "https://example.com/your-style-lora.safetensors",
"scale": 1
}
]
}),
}).then((r) => r.json())
while (true) {
const { data: result } = await fetch(`https://api.wavespeed.ai/api/v3/predictions/${task.id}/result`, { headers }).then((r) => r.json())
if (result.status === 'completed') { console.log(result.outputs); break }
if (result.status === 'failed') throw new Error(result.error)
await new Promise((r) => setTimeout(r, 2000))
}WaveSpeedAI vs Civitai
| WaveSpeedAI | Civitai | |
|---|---|---|
| Billing | Pay per run in US dollars, with the price shown on every model. No subscription | Buzz credits, bought in packs or included in monthly memberships |
| Running LoRAs | LoRA endpoints for FLUX, Qwen Image, Z-Image, Wan, LTX and MiniMax H3 | On-site generator for community checkpoints and LoRAs |
| LoRA training | Image and video LoRA trainers, in the browser or by API | On-site trainer for SD 1.5, SDXL, Flux and some video models |
| Image and video models | Seedream, Nano Banana, GPT Image, Seedance, Wan, Kling, MiniMax H3 and more | Community models plus some third-party image and video models |
| API | Every model and trainer by REST, CLI or MCP, at the same price as in the browser | Catalog API, plus a developer API for generation metered in Buzz |
FAQs
Is WaveSpeedAI a good Civitai alternative?
Yes, if you want to run and train LoRAs and generate images and video without a subscription. LoRA endpoints cover FLUX, Qwen Image, Z-Image, Wan, LTX and MiniMax H3, trainers turn your own images into new LoRAs, and the newest image and video models are on the same account. You pay per run, and everything is available by API.
Can I use a LoRA I already have?
Yes. LoRA endpoints take the URL of your LoRA file and a strength, and many let you combine several LoRAs in one run. Pick the endpoint for the base model the LoRA was trained on: a FLUX.1 LoRA on the FLUX.1 endpoint, a Wan 2.2 LoRA on a Wan 2.2 endpoint.
How do I train a LoRA?
Open a LoRA trainer, upload a zip of images (caption files are optional), set a trigger word and the number of steps, and start. The price is shown before you run and scales with the steps. When training finishes, you get a LoRA file to use on the matching LoRA endpoint.
Can I train and run video LoRAs?
Yes. The Wan 2.2 and LTX-2 trainers produce video LoRAs that run on the matching Wan and LTX LoRA endpoints, and MiniMax H3 has LoRA endpoints for text-to-video and image-to-video too.
Do I need a subscription?
No. Add credits when you need them and spend them on any model or trainer. No monthly fees, nothing to cancel.
Can I try it for free?
Yes. Eligible new accounts get $1 in free credits, and you can sign up without a credit card. Some premium models may not be available with trial credits.
Is there an API?
Yes. Every LoRA endpoint, trainer and generator is one REST call away with a single key, at the same per-run price shown on its model page. Pass your LoRAs in the request, and use the CLI or MCP server to hand the same models to your agents.
Bring your LoRAs. Make something new.
$1 in free credits for eligible new accounts. No card needed.