Text to image in about two seconds with eight-step distilled model
AI image and video upscaler with 14 specialized models, up to 8K output
Stage
Not given
Not given
Founded
Not given
Not given
Based in
Not given
Not given
Pricing model
Freemium
Freemium
Pricing
Free credits available; paid annual plans with 50% launch discount: Basic $9.9/month (800 credits/month, 9600 credits/year), Professional $19.9/month (2000 credits/month, 24000 credits/year), Enterprise $44.9/month (6000 credits/month, 72000 credits/year). Monthly plans also available. Per-render co
Free: 5 upscales per month at 2K resolution, watermarked. Creator: $9/month for 50 images and 20 videos up to 8K. Pro: $19/month for 100 images and 30 videos. Studio: $39/month for 250 upscales and 90 videos. Business: $49/month for 500 upscales and 150 videos with API access. Annual plans save 20%.
Key features
Eight-step sampling produces 1024px drafts in roughly two seconds
Text to image and image to image workflows in one workspace
Support for LoRA attachments with no extra render cost
Resolution from small exploratory sizes up to 2K
Open weights published under custom licence permitting commercial use
Photographic base behaviour from 12B model trained from scratch
LoRAs behave as bipolar sliders for intensity control mid-series
Browser-based workspace with no GPU or local install required
Same 12B backbone as standard Krea 2 via trajectory distillation
Unlimited exports in PNG, JPG and WebP formats
14 specialized AI models for different upscaling jobs
Upscale images and videos up to 8K resolution
Resumable uploads supporting files up to 2GB
REST API for programmatic upscaling
Most jobs complete in under 30 seconds
Watermark-free results on Pro and Business plans
Commercial license included on paid plans
What makes it different
Two-second generation speed enables iteration-heavy workflows instead of single-frame rendering
Open weights allow running the same model locally without vendor lock-in
Trajectory distillation preserves detail of full 12B model at eight-step speed
14 transparent models, each labeled whether it invents detail or only recovers existing information
Unified cloud platform handling both images and video without separate tools
Batch processing and API access for production workflows
No file training: original media is never used for model improvement