Use media models directly.
Call optimized, auto-scaled, production-ready image and video models. Send JSON, get a media URL.
- ✓ Get started fast
- ✓ Pay by output or usage
- ✓ No GPU management required
A generative media platform built for products and Agents.
Use production-ready image and video models, deploy your own models, or run them on dedicated compute when your product needs more control.
Image & Video Models · workflow
01Select a visual task
02Keep the interface direct
03Carry it into an Agent
- Media models
- 100+
- Ways to use
- 3
- Shared credit & records
- 1
Start from existing assets or models.
Different levels of control over models, code, and infrastructure suit different products.
Deploy your own model.
Bring model weights, a container, or an existing service. SandBase manages GPU, queues, scaling, and observability.
- ✓ Control code and weights
- ✓ Auto scale to zero
- ✓ Pay by runtime
Continuous use of dedicated GPU.
Ideal for training, fine-tuning, batch processing, and long-running workloads requiring stable hardware access.
- ✓ Full SSH control
- ✓ Fixed hourly pricing
- ✓ Single or multi GPU
View one category at a time.
Select a classification dimension first, then choose a specific filter; irrelevant options stay out of the way while the API call remains consistent.
nano-banana-2/edit
Production-ready generative media model API for creative workflows.
Try it →nano-banana-pro/edit
Production-ready generative media model API for creative workflows.
Try it →openai/gpt-image-2/edit
Production-ready generative media model API for creative workflows.
Try it →flux/schnell
Production-ready generative media model API for creative workflows.
Try it →nano-banana-2
Production-ready generative media model API for creative workflows.
Try it →Call a model, deploy a model, or manage dedicated compute?
Choose by what you need to control, how the workload scales, and how usage should be billed.
| Product | What you control | Scaling method | Billing | Best suited for |
|---|---|---|---|---|
| Model APIs | Input and model selection | Automatic | Per output / usage | Product features, Agent media capabilities |
| Serverless | Code, weights, container | Automatic, from zero | By runtime | Private models, custom pipelines |
| Compute | Full GPU instances | Manual, continuous runtime | Fixed hourly pricing | Training, fine-tuning, batch processing |
Infrastructure is not the end; deliverable work is.
Most customers start with Model APIs. Verified media capabilities can enter the product directly, or be equipped with language models, APIs, and Skills for Agents.
01
Select supply
Model API · Serverless · Compute
02
Verify capabilities
Quality · Latency · Cost
03
Delivery
Product API · Media workflow · Agent
Run results and cost return to the next model selection ↺
Start generating in three steps.
Create an account, add one shared balance, and use the same API key across image, video, and language models.
- 01
Create account
Register for SandBase; your personal workspace is created automatically. Invite team members when collaboration is needed.
You · Team
- 02
Top up shared credits
Start from ¥50. The same shared credit works across image, video, and language models.
¥50 · Image · Video · Language
- 03
Get API key
Create one SandBase API key, select a media model, and receive a deliverable media URL.
POST /v1/media/generate
A visual model is only the beginning.
Bring the result into an AI tool or connect it to APIs that turn media generation into repeatable work.