Astria.ai
Astria.ai.ai is an AI tool that provides custom image creation capabilities with high-quality fine-tuning, virtual try-on, and motion generation features.
Last verified:
What is Astria.ai?
Astria.ai is a generative AI image platform that provides a fine-tuning API for creating custom, personalized AI image models. The service allows developers and businesses to adapt pre-trained generative image models like Flux, Stable Diffusion 1.5 (SD1.5), and Stable Diffusion XL (SDXL) to generate studio-quality images of specific subjects or styles. Through a single API call, users can perform fine-tuning, use batch predefined prompts, upscale images, apply face correction, and access many other generative imaging features.
Key features include AI-generated professional headshots, custom avatars and cartoons, virtual try-on with Flux using garment reference images, pet image generation, product shot customization with background and lighting adjustments, interior design visualization, generative filters for artistic effects like line-art and oil painting, super-resolution upscaling, and video generation at 24 frames per second. The platform supports FaceID/adapter alternatives alongside Dreambooth fine-tuning, offers Checkpoint and LoRAs fine-tuning types for both SDXL and SD1.5, and provides a model library where users can use any open model or import their own as a baseline.
Astria.ai is designed for developers building AI photography services, mobile app creators embedding users in images, corporate businesses needing professional headshots, pet stores wanting to monetize with AI-generated pet images, e-commerce businesses requiring product shots, interior designers redecorating rooms, and anyone needing generative filters while preserving subject identity. The platform offers unique infrastructure adjusted for fine-tuning with auto-scaling, no queues, fast processing, and handles jobs in parallel without wait time.
Astria.ai pricing
Pricing model: Freemium
Astria uses a pay-as-you-go freemium pricing model with no free tier for ongoing use. New users receive $20 free credits upon generating an API key. Pricing: $1.50 per fine-tune (AI model creation), $0.10 per prompt (generates 8 images by default), $0.40 per 100 frames for video generation (24fps), and $0.50 per model per month for extended storage beyond 30 days. Minimum payment is $5.00 added to account balance. Credits must be used within one year or they expire. Super-resolution or face correction prompts default to 4 images. Pricing is billed monthly or yearly with pay monthly or pay yearly commitment options.
Astria.ai pros
- Single API call for fine-tuning, batch prompts, upscaling, and face correction
- Supports Flux, Stable Diffusion 1.5, and Stable Diffusion XL models
- Generates studio-quality AI professional headshots
- Creates custom avatars and cartoons with consistent characters
- Virtual try-on feature with garment reference images using Flux
- No DevOps required with auto-scaling infrastructure
- No queues with fast processing and parallel job handling
- Generative filters preserve subject identity and similarity
- Super-resolution option for higher resolution pictures
- Download CKPT files for use with Automatic1111 and ComfyUI
- Video generation at 24 frames per second
- $20 free credits upon generating API key
- Model library with open models and custom import options
- FaceID/adapter alternative requires less processing time
- Both Checkpoint and LoRAs fine-tuning types available
- Interior design from sketch to reality capability
- Commercial use allowed without additional licensing requirements
- Handles peak loads for biggest companies and app-store applications
Astria.ai cons
- Not free - requires payment for model creation and image generation
- $1.50 per fine-tune can add up for multiple model iterations
- $0.10 per prompt cost for generating 8 images
- Models deleted after 30 days unless storage extended for $0.50/month
- $0.40 per 100 frames for video generation is expensive
- Minimum payment amount of $5.00 required
- Credits expire within one year of purchase
- Model very sensitive to JPEG artifacts in uploaded images
- Extended model storage requires billing page configuration
- Training couples or multiple people harder for variation in long-shots
Frequently asked questions about Astria.ai
What is fine-tuning?
Fine-tuning is the process of adapting a pre-trained generative image model like Flux, Stable Diffusion 1.5 (SD1.5), or Stable Diffusion XL (SDXL) to create personalized results for specific subjects or styles. Astria's fine-tuning API allows generating studio-quality images of a subject, including AI-generated professional headshots, as well as cartoons and avatars.
What image size should I upload and is there a size limit?
There is no hard limit on image uploads, though the service has limits against JPEG bombs. A reasonable limit is 3MB per image. Image size should be at least 512x512 pixels since models train on 512x512, but 1024x1024 or 2048x2044 is a good target resize to allow headroom for face-cropping. Make sure compression does not introduce JPEG artifacts as the model is very sensitive to them.
Why isn't Astria free?
For each model created using training images, Astria creates an AI model using high-end GPU machines and then serves those models to allow image generation. This process is expensive due to the computational resources required.
What is Flux?
Flux is a family of generative image models created by Black Forest Labs that allows a new level of text-to-image generation. FLUX.1 launched in August 2024 with 12 billion parameters, delivering exceptional image fidelity and controllability. It uses flow matching and Denoising Image Transformer (DIT) architecture, outperforming SDXL and SD1.5 in image quality and prompt adherence.
Where can I see my generated images?
Check out your tunes page and click on the tune thumbnail to see your results. To download, go to https://www.astria.ai/tunes, click on the tune thumbnail, and hit download to save the pictures of each prompt.
Can I get higher resolution pictures?
Yes, toggle the Super-resolution option in the prompt Advanced section to get higher resolution pictures.
Can I download CKPT files and how do I use them?
Once the fine-tune is done processing, you will see TAR and CKPT buttons at the top of the fine-tune page. You can use the CKPT file with software like Automatic1111 and ComfyUI for local image generation.
What happens to my model, training pictures, and image generations?
The entire fine-tune including CKPT file, model, training pictures, generated images, and prompts is kept for 1 month, after which everything is deleted. You can delete it at any time by clicking the delete button in the fine-tunes page.
Can I extend model storage beyond 30 days?
Yes, you can select to extend model storage in the billing page. Extended storage costs $0.50 per model per month.
Can I train a model for more than one person?
Yes, you can train a fine-tune using class_name=couple and get amazing results. Make sure that the couple appears in each of the training images together. However, it might be harder for the model to create more variation and produce long-shots and complex scenes in this case.