Choosing between four Venus image tools, and writing a prompt for one track
Reality 1 Lightning, Reality 1, Reality 1 Pro and Super 2 make images at different speeds and levels of control. Tempo 1 makes one track per five hours. How to choose, and what to put in each prompt.
- Published
- 3 min
- To read
- Venus
- Project
VenusMCP has four image tools and one music tool, and they are not interchangeable. Each has a different speed, a different amount of control and a different limit. Choosing the right one at the start saves a regeneration, and for two of them a regeneration is not available for a while.
Four image tools
| Tool | Best for | Speed |
|---|---|---|
| Reality 1 Lightning | A simple, well known subject where speed matters more than control | A few seconds |
| Reality 1 | Quick or loose images where the details are not critical | Fast |
| Reality 1 Pro | An image where the details matter and the result has to be right the first time | Slower |
| Super 2 | Dramatic scenes with strong light and text that has to be readable | Up to a minute |
Reality 1 Pro is the unusual one. A thinking step reads your idea and the genre you want, expands them into a complete prompt that leaves the image model nothing to guess, and only then renders. It can ask up to three questions first when something decides the picture and your request did not say. Answering them is quicker than regenerating.
Super 2 is limited to three images per hour for each IP address, so it is worth writing the prompt before you send it.
What goes in an image prompt
The same few things decide most images, whichever tool renders them.
- The subject, concretely
- A weathered fishing boat is better than a boat. Count things that need counting.
- The setting and the light
- Time of day, direction of the light and weather change a picture more than any adjective about style.
- The framing
- Close up, wide, from above, from the water. Say where the subject sits in the frame if it matters.
- A medium, when you want one
- A photograph, a pencil drawing and a render are different pictures of the same boat.
- Text, spelled out
- If words must appear, give the exact words in quotation marks and say where they go. Leave text out entirely when it does not need to be there, because stray lettering is the most common flaw in an otherwise good image.
The defaults lean toward photographic realism and restrained colour. Ask for something else directly if you want it. The defaults also avoid imitating a named artist.
One track, once
Tempo 1 writes one short track as an mp3. It is limited to one track per IP address every five hours, and a refused request should not be retried. The prompt has to be right before it is sent.
A prompt that works describes the track as a brief to a musician:
- Genre and feel
- Slow, warm, late night piano jazz, or tight electronic with a four on the floor kick.
- Tempo
- A number of beats per minute, or a word like unhurried.
- Instruments
- Name two or three. Naming ten gives a muddle.
- Structure
- Say if it builds, drops or loops. A short track with a clear arc beats one that tries to do everything.
- Lyrics or none
- Say instrumental if you do not want a voice.
A mockup before you build
If the picture you want is an interface and not a photograph, use Design 1. It returns a high fidelity mockup together with design notes: palette, type scale, spacing and radii. Those notes are what you build the real screen from, so the mockup is a starting specification and not only an image.
The full list of VenusMCP tools is in the guide, and Trail 1 covers planning work that takes more than one session. The server is at venusmcp.caissonai.co.
Written by Caisson AI. Questions or corrections go to hello@caissonai.co.