Finished result
This nine-second Adesso concept shows the workflow in practice. Bloom supplied the brand system; Seedance rendered the video. The walkthrough below now generates video through Bloom MCP.What you need
- An agent connected to Bloom through MCP.
- A ready brand in Bloom.
This feature is in beta. During beta, endpoint schemas and behavior may change
or be removed.
Run it with your agent
Connect Bloom to your agent
Setup for Codex, Claude, ChatGPT, Cursor, or any MCP-compatible agent.
How it works
1
Load the brand system
Retrieve the complete active Brand Skill from Bloom before making creative
decisions.
2
Distill it for video
Turn the relevant guidance into a video request. The agent reads the full
Skill, then sends only the focused prompt and useful assets to Bloom for
video generation. Raw font files are not useful inputs for a video model. If
typography matters, use rendered references.
3
Render through Bloom
Use Seedance 2.5 through Bloom to render the clip from the request and any
chosen references. The Adesso example used a reference image; other concepts
may not need one. Keep checking progress until the clip is ready to review,
or report a failure if generation fails.
4
Review and refine
Compare the video with the Brand Skill. If it misses the request, adjust the
prompt or references and render again as needed.
What each system did
- Bloom maintained the brand context and supplied the active Brand Skill.
- The agent chose the concept, prompt, and assets, and handled review.
- Seedance rendered the Adesso example. The current walkthrough uses video generation through Bloom MCP.
Before you publish
- Check brand fit against the Brand Skill.
- Look for malformed logos, incorrect text, and product or subject deformation.
- Watch for visual artifacts and check framing throughout the clip.
- Listen to the sound and check that it fits the intended use.