Image generation
Text-to-image in 16:9, 9:16, or 1:1 - up to 4 variations per call. Condition on one reference image, or several labelled ones for character and style consistency.
Generate media from any AI agent
Connect Dolphin to Claude, Cursor, or any MCP client. Images, video, long-form video, and speech - one URL away.
https://mcp.trydolphin.ai/mcp
- Claude
- Gemini
- OpenAI
- Cursor
- Perplexity
Setup
Connect once, generate from any conversation.
Open Connector Settings
In the Claude app or on claude.ai, go to Settings → Connectors and choose Add custom connector.
Open Claude settingsAdd custom connector
Click Add Custom connector and paste the URL - https://mcp.trydolphin.ai/mcp
Sign in and start creating
Authenticate with your Dolphin Studio account. Ask Claude to generate an image or video.
https://mcp.trydolphin.ai/mcp
Ask, and Dolphin Builds It - Right There in Chat.
Your AI agent handles everything - model selection, polling, delivery.
Generate video right in chat
Generate a cinematic 16:9 video of a drone shot over a misty forest at sunrise, about 6 seconds.
Task registered with an ETA. The agent returns the finished clip as a downloadable URL.
One agent,
every tool in Dolphin
Everything you need to generate media programmatically.
Image generation
Text-to-image in 16:9, 9:16, or 1:1 - up to 4 variations per call. Condition on one reference image, or several labelled ones for character and style consistency.
Video generation
Text-to-video and image-to-video with first/last frame control. 480p to 1080p. Most models run 4–10s, a few up to 15s.
Long-form video
One script, one reference image, one talking video - 30 seconds to several minutes, speech generated automatically. Always 9:16.
Text to speech
Any text, spoken - pick a voice from a full catalogue with previews, accents, and use-case labels.
Reference-driven control
Guide generation with labelled reference media - image for most models, plus video and audio on Seedance 2.0.
Fully async
Register a task, get an ETA, poll for status - or skip the polling and get the finished asset back in one call.
Credit billing
Same credits, same workspace balance as Dolphin Studio. Check the cost upfront, or see it returned with the result.
Saved defaults
Pin a default model and voice once - they persist across conversations, so calls that skip a model just work.
Secure authentication
OAuth for web agents, API key for CLI. No credentials to manage in your prompts.
Latest AI Models, all in one place
FAQ
Via MCP (Model Context Protocol) — an open standard that lets AI agents call external tools. Add the server URL, authenticate, and your agent gets access to all generation tools.
Claude (web + desktop), Claude Code, Cursor, and any MCP-compatible client. Web agents sign in with your Dolphin account; CLI tools use an API key.
For Claude.ai — no. OAuth handles everything automatically. For Cursor or Claude Code, generate a key from Dolphin Studio under Settings → API Keys.
Same credit system as Dolphin Studio. Cost varies by model, resolution and duration. Your workspace credits work through any connected agent, and your agent can quote the cost of a job before running it.
Every job returns an ETA when it's registered, and your agent polls automatically. Images are usually the quickest; short video typically takes a few minutes and can sit queued before it starts; long-form video runs 10–20 minutes. Times vary by model, resolution and current load.
Images in 16:9, 9:16 or 1:1, up to 4 variations, with single or multi-reference conditioning. Short video from text or images, 480p–1080p, 4–15s depending on the model, with first/last frame control. Long-form 9:16 video from a script and a reference image. And speech from text in any of the available voices.
Yes. Pin a preferred image model, video model and voice once, and any call that omits them uses your choice. The preference persists across conversations until you change it.
