Welcome to Lorvara.
We built the platform to serve as a frontier model harness that lives between your Claude or ChatGPT chats and your pro production tools like Resolve, Premiere or Final Cut.
In a nutshell: take the creative briefs you’re already writing in your LLM, put them into Lorvara, and they get split into a manageable set of scenes, shots and prompts that manage your reference images across models like Seedance 2.5.
The unit you work in is a Key Prompt, and it’s simple. It’s a combination of:
- A natural language prompt, your base prompt
- The MediaGen model selected for the media
- Your parameters for that model
- Named reference media stored in the prompt’s library
Lorvara takes your creative concept, structured with your LLM, and sets up the prompts for you to refine and generate the media you need with precision and consistency.
Why briefs
Creators are routinely using Claude to manage their projects. Lorvara can take that input, in the form of briefs, to create the characters, scenes, props and states that are the staple of professional production.
Think of Lorvara as the human interface that’s critical to the creative process. Do all of your story and shot planning in your LLM, then when it comes time to generate, Lorvara handles the final prompts and organizes the assets, which is exactly when you want help, because that’s when you start paying for expensive frontier models.
The shape of it:
LLM brief → project → shot list (up to 40 shots per pass) → batch prompts → prompt agent → references → generate
Claude is built in
The best image, video and audio results will always need some manual review and iteration. Lorvara makes that workflow more fluid with Claude built right in.
It’s an expensive LLM to work with, so we’ve narrowed what it has to figure out. Behind it is our own model record: three years of measuring how these mediagen models actually behave rather than what their spec sheets claim.
You’ll want to use Claude in Lorvara. There are three agents, but the Studio’s Prompt Agent is where you’ll spend most of your time. It costs credits, and it will save you time and write better prompts than Claude alone can. It also reads Skills. Try ours, or bring your own.
Where this came from
The agents use the Key Prompt system I developed back in 2023. I started a project to create prompts for Midjourney. It was ugly, the UX was terrible, but it produced some great outputs on MJ’s 6.1 and 7 models.
I launched Key Prompt Manager and Key Prompt Ninja over the next few years. Along the way I got a lot of great feedback, and I’ve turned that into a next-generation prompt builder, designed for how the best creators using the latest MediaGen models actually work.
The models
Lorvara is largely built around final productions using Seedance models, but it supports 46 models across image, video, audio and upscaling. Other popular models include Kling, Nano Banana Pro, GPT Image 2.5 and the latest from Google and Grok.
Midjourney
I know a lot of you are Midjourney fans. I am too. Unfortunately they don’t provide an API. That doesn’t stop us.
Lorvara provides full MJ integration into your projects, with full model version, parameter and reference support. Our Midjourney prompts carry your references across as public CloudFront URLs. Style, character and image prompts all go, each on the right flag for the version you picked, so they work the moment you paste them.
You can also copy image links from Discord to import straight into your Lorvara library.
For teams
For Directors and Producers running a team, Lorvara has a real collaboration system.
You can run a team of 5 or 10, including you, sharing projects and asset libraries. You get more project control with built-in prompt status and approvals, so the team can see what’s still being worked on and what’s signed off. The team leader manages a credits bank the whole team draws from, and you can see creator activity across the current period.
Getting started
Sign up and choose a plan. Every plan includes monthly credits for the Prompt Agent and for generation.
From there you can start either way. Open the Studio and make one picture in six steps, nothing set up first. Or bring a brief you’ve already written in your LLM and let it become a shot list.
This is just the beginning, so please tell me how it can work better for you.