Talking to Vizzy
How to write prompts that get the render you want, reference uploaded images, and change settings like aspect ratio, model, or duration just by chatting.
Vizzy understands plain language. You don't need prompt-engineering tricks or a syntax to learn — but a few habits get you to the right result faster.
Describe the outcome, not the settings
Focus on what the final image should look and feel like. Vizzy translates that into the right model, parameters, and detailed prompt internally.
Less effective: "render mode, 4k, photorealistic, octane" More effective: "A photorealistic dusk shot of this house with warm interior lights on, a wet driveway, and soft rim light on the roofline."
Useful things to include:
- Lighting and time of day — "golden hour", "overcast", "warm evening interior".
- Materials and finishes — "matte black cladding", "oak floor", "brushed brass".
- Mood and medium — "editorial architecture photo", "watercolor concept", "blueprint".
- What to keep or change — "keep the layout, change the furniture to mid-century".
Referencing your images
When you've attached or already generated more than one image, type @ in the message box. A dropdown of your images appears — pick one to drop a reference to it straight into your prompt. That's how you tell Vizzy which image is the subject and which is a style or material reference.
Use @the base sketch and apply the style of @the moodboard.
You don't have to use @ — you can also just describe the images in words ("the sketch", "the moodboard") and Vizzy will work out which is which. When the roles are ambiguous, it asks before generating.
Changing settings by chatting
Every parameter can be set in conversation — no need to open the advanced controls. Vizzy updates the setting, confirms it inline, and the form controls update to match.
| Say something like | What changes |
|---|---|
| "Change to 16:9 widescreen" | Aspect ratio |
| "Make it portrait" | Aspect ratio |
| "Use the pro model" | Generation model |
| "Higher resolution please" | Output resolution |
| "Make the video 5 seconds" | Video duration |
| "Add sound" / "loop it" | Video audio / loop |
After a settings change, Vizzy will often offer to regenerate with the new setting — say yes, or keep chatting.
Refining versus restarting
You almost never need to start over. Replying in the same thread keeps the context of your last render:
- Refine — a small delta: "warmer", "remove the car", "more plants".
- Restyle — same subject, new look: "now as a watercolor", "blueprint version".
- Explore — when you're not sure yet: "show me a few directions for this lobby" and Vizzy proposes options before committing (see Batch).
Pinning skills
If you consistently work in one discipline, you can pin the relevant persona skills so they're always loaded — useful when your prompts are short and you want Vizzy to assume your context every time.