Medium Shot — AI shot size prompt
Medium Shot is a framing you can specify instead of leaving it to chance. Here is the phrase to use, what it changes, and a shortlist of models you can try it with.
The prompt phrase
Medium Shot
medium shot framed from the waist upShot size controls how much subject and environment the frame includes. Naming it makes that composition goal explicit.
A worked example
A chef plating a dish in a busy restaurant kitchen, medium shot framed from the waist upYour subject comes first, then the technique, then anything else. Models weight the front of a prompt most heavily, so keep the subject there and let the shot size modify it.
Models to try it with
| Model | Credits | What it is for |
|---|---|---|
| Flux Schnell | 1 | instant drafts, lowest cost |
| Z-Image Turbo | 1 | ultra-fast drafts and variations |
| Flux Dev | 2 | good quality without the top price |
| Nano Banana | 2 | follows instructions most closely |
| Recraft v3 | 3 | logos, icons and clean vector art |
| Seedream 4.5 | 3 | photoreal, and gets text right |
| Flux 2 | 4 | newest model, most realistic results |
| Flux Pro | 4 | sharpest detail of the Flux family |
In video
You can also try the same phrase in a video prompt alongside a camera move:
| Model | Credits | Native audio |
|---|---|---|
| Veo 3.1 Lite | 6 | Yes |
| Veo 3.1 | 22 | Yes |
| Veo 3.1 Fast | 12 | Yes |
| Kling 2.6 Pro | 11 | Yes |
| Kling 3 | 12 | Yes |
Related shot sizes
- Extreme Close-Up shot size
- Close-Up shot size
- Cowboy Shot shot size
- Wide Shot shot size
- Extreme Wide shot size
- Two Shot shot size
- Insert / Detail shot size
Common questions
What exactly do I put in the prompt?
The phrase above, word for word: "medium shot framed from the waist up". Put it after your subject, separated by a comma.
Which model should I try first?
The table lists 8 models available for trying this prompt. Flux Schnell is the cheapest starting point at 1 credits, then move up a tier once the shot works.
Can I combine it with other shot sizes?
Yes. Start with two or three compatible instructions and compare the output; conflicting directions may be interpreted differently by each model.