Modelflare Blog

Castaly: an AI video and image studio built on Modelflare

Castaly is a browser studio for AI image and video generation. Models are compared by specification and price, generations are paid from a wallet shared with Modelflare, and every model is also available through the Modelflare API.

Orbital ring city, generated with gemini-3-pro-image

Cover: "Orbital ring city", generated in Castaly with gemini-3-pro-image (2K, 16:9), cropped to 1200x630.


Castaly is an AI video and image studio. It generates video and images with leading models, shows their specifications and prices side by side, and exposes each model's parameters directly. It runs on the Modelflare gateway, so the models used in the studio are the same ones available from code.

Choosing a model

World-building, character work, and action shots each depend on different model strengths. Castaly organizes the choice by task and then compares specification and cost across candidates. Each result keeps the model and specification it was generated with and is stored in the user's library. The homepage's "Make any video" section shows eight examples, each a real output from a model Castaly runs; opening one starts a new generation from the same prompt.

A humpback whale floating above a wheat field "Whale over wheat", gpt-image-2.5-flare, 1536 × 1024.

Available models

Castaly uses the image and video models in the Modelflare catalog, which the documentation lists as 14 image models and 28 video models. The workbench lists only the models enabled for the signed-in account. The general-purpose models include:

Type Family Model IDs
Image OpenAI Images gpt-image-2.5-sunburst, gpt-image-2.5-flare, gpt-image-2
Image Gemini Image gemini-3-pro-image, gemini-3.1-flash-image, gemini-3.1-flash-lite-image, gemini-2.5-flash-image
Image Wan, Seedream wan2.7-image, wan2.7-image-pro, seedream-5-0-lite-260128
Video Seedance seedance-2.5, seedance-2.0, seedance-2.0-fast, seedance-2.0-mini
Video Wan wan3.0-video-prime, wan3.0-video, wan2.7-t2v, wan2.7-i2v, wan2.7-r2v
Video Kling kling-3.0, kling-3.0-omni, kling-3.0-turbo, kling-o1, kling-2.6, kling-2.5-turbo
Video MiniMax, HappyHorse MiniMax-H3, happyhorse-1.1-t2v, happyhorse-1.1-i2v, happyhorse-1.1-r2v

The studio itself is for images and video. Language models such as GPT, Claude, and Grok are used through the Modelflare API and billed per million tokens, with separate input and output rates.

An empty room whose window holds back an ocean "An ocean behind the glass", Seedream 5.0 Lite, 2560 × 1440.

Workflow

Users sign in with a Modelflare account; the Castaly account, wallet, and usage history are the Modelflare ones. In the workbench, choose image or video, then a model. The price per image, per second, or per task appears next to the selector. Write a prompt, optionally attach a reference file or a library asset, and set the aspect ratio, resolution, quantity, or duration in the settings panel.

Each task is settled in USD from the shared wallet. The estimate and the actual cost of every task are shown under Creations. Finished images can be saved to the library and reused as references, and results can be published to the community with or without their prompt and settings. Published work can be withdrawn at any time from the My page.

A rainy Tokyo alley at night with neon reflections "Rainy Tokyo alley", wan2.7-image-pro, 2K.

Pricing

Castaly charges per generation, with no subscription and no minimum. Each image or video is billed on actual model usage; browsing, comparing, and writing prompts are free. Resolution, duration, size, and quality determine the rate, and the price updates before generation. Funds go into a single wallet that also pays for direct API calls, and the balance does not expire. Current per-model prices are published on the pricing page.

Community

The Community page shows generated video and images together with the prompts and settings that produced them. Any prompt can be copied, and any work can be remixed in the studio.

A salt flat whose horizon peels up to reveal night "The horizon, lifted", Seedream 5.0 Lite, 2560 × 1440.

Calling the same models from code

The Modelflare gateway handles model access, balance, and usage. To call the same models from code, get a key from the My page, then either install the Modelflare Media Skill or copy the request from a workbench result, and set the key in the environment. A documented Seedance request: Accounts created through Castaly already have a key named Castaly that routes across every group enabled for the account. Each workbench generation includes a "Reproduce this in your own agent" panel containing the install message, the skill package, and a task brief.

curl https://modelflare.dev/v1/videos \
  -H "Authorization: Bearer YOUR_MODELFLARE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "seedance-2.5",
    "content": [{"type": "text", "text": "A cat walks through a sunlit room"}],
    "resolution": "720p",
    "duration": 5,
    "ratio": "16:9"
  }'

Enterprise

Castaly Enterprise provides higher limits, priority support, and volume pricing for teams generating video and images through Castaly and the Modelflare API. Contact business@modelflare.dev.

References

Studio: castaly.modelflare.dev. Documentation: Castaly docs. Pricing: Castaly pricing. Modelflare on Hugging Face: organization, api-recipes, coding-agents. Business inquiries: business@modelflare.dev.

← All articles