Hedra Agent 2 builds a 30-second video pipeline from one prompt

• Hedra Agent 2 turned “What if AI ran a restaurant?” into research, a brief, script, characters, environment, and a cinematic video inside one workflow, according to AIwithSynthia's demo.
• The behind-the-scenes capture was the workflow story: the 30-second screen recording attached to AIwithSynthia's follow-up kept research, scripting, visuals, and production inside one Space.
• The workflow lands because it removes tool-hopping, with AIwithSynthia's reply saying the work happened under one Space and another reply calling it a single-tool run.
• Creator reaction framed the demo as solo-studio leverage: one reply called it “one person studio energy,” while AIwithSynthia answered another viewer with “Future seems very crazy.”

Hedra's overview docs call Agent an “agentic workspace for research, visual inference, and creative production,” and its Spaces docs define a Space as the place where conversation, canvas, references, and outputs stay together. The current Hedra homepage says “Create with Agent 2” and pairs Agent with Skills, MCP, multiplayer Spaces, and a model stack; MiniMax also announced that its video and text-to-speech models are integrated with Hedra.

AIwithSynthia started with one prompt: “What if AI ran a restaurant?” The demo then broke the video job into five visible stages:

1. Concept research.
2. Brief building.
3. Script writing.
4. Character and environment creation.
5. Cinematic video production.

A reply from AIwithSynthia said the concept research worked “very well.” For AI filmmakers, the useful part is the handoff collapse: strategy, writing, art direction, and generation sit in one run instead of four browser tabs.

The follow-up attached a 30-second screen recording of the build process. In AIwithSynthia's follow-up, the claimed sequence was research, brief, script, visuals, and production without jumping between multiple tools.

Two replies compressed the workflow into creator shorthand:

• “All work done under one space,” as AIwithSynthia put it.
• “This worked with a single tool,” as AIwithSynthia wrote in another reply.

Hedra's Spaces docs describe a Space as the working context for a project. It contains:

• Conversation: the brief, follow-up instructions, decisions, and agent results.
• Canvas: generated, uploaded, and collected media arranged visually.
• References: assets and Elements used for identity, style, or product consistency.
• Project context: completed work available for the next request.

Hedra's May upgrade post says the canvas can hold text blocks, images, videos, and research, and that the agent reads those pieces as working context. That matches the restaurant demo's appeal: the brief lives next to the script, and the script lives next to the video.

Hedra's Skills and connectors docs turn repeatable creative workflows into saved instructions. Built-in examples include:

• Storyboards.
• UGC.
• Product showcases.
• Character sheets.
• Restyling.
• Moodboards.
• Campaign briefs.
• Competitor scans.

Connectors let the agent pull outside context into a Space, according to the same docs. Hedra's upgrade post also says the agent can read and write across connected systems through MCP, run code in a sandbox, use Exa for web research, and choose media models for video, image, and audio outputs.

The demo thread ended with a direct Hedra link. Hedra's Agent page says Agent runs on the same credit system as the rest of the platform: pay for generated work, not access to the tool.

The same page lists individual plans at $15 per month for 1,500 credits, $30 per month for 5,400 credits, and $75 per month for 14,400 credits. It also says Agent can select from 14+ AI models, carry context across the conversation, adapt content across platforms, and export high-resolution output or keep building inside the project.

The model stack is moving too. MiniMax's announcement says its video and multilingual TTS models are now integrated with Hedra for video generation and speech output.