How to write Nano Banana 2.1 prompts for film work: a reusable structure, three worked examples, Camera Angle Control tips and the common mistakes to avoid.
Nano Banana 2.1 is Google's image model and the default image model in Cinemagiq's Generation Studio. For film work it covers most of the stills a production needs: location concept art, storyboard frames, character sheets, posters, title cards, and the first frames you later animate. It reads plain, descriptive sentences, so the craft is less about magic words and more about deciding what the frame is before you type.
This guide gives you a prompt structure, three worked examples, and the settings that matter.
The prompt is the only required input. Everything else is optional and set around it:
Lead with what kind of image this is ("concept art", "film still", "character sheet"), so every detail after it is read in that context. Then describe the content. Put framing last, or leave it to Camera Angle Control.
[Image type or medium]: [subject with two or three concrete details] [doing what] in [place], [light and time of day], [mood or color], [shot size or framing].
Fill each slot once. Leave a slot out rather than pad it. If you set an angle with Camera Angle Control, keep the angle out of the text so the two never disagree.
Made in Cinemagiq. The exact prompt is under each one.
PromptLocation concept art, matte painting: a floating island village above a sea of clouds at sunrise, wooden rope bridges between the rocks, waterfalls falling into the mist, golden rim light, epic wide establishing shot.
PromptA knight in battered silver armor standing in a burning castle courtyard, embers drifting in the air, dramatic backlight, cinematic film still.
PromptA knight in battered silver armor standing in a burning castle courtyard, embers drifting in the air, dramatic backlight, cinematic film still.
Example 1: location concept art. The prompt opens with the deliverable, "Location concept art, matte painting:", then lists the set pieces a production designer would ask for: "wooden rope bridges between the rocks, waterfalls falling into the mist". Light gets one precise phrase, "golden rim light", and the closing words, "epic wide establishing shot", set the framing. Every slot of the template is used once, and nothing is padded.
Example 2: Low Angle. This prompt describes a character, a place and a light source ("dramatic backlight") and labels itself a "cinematic film still", but it says nothing about where the camera is. That job goes to Camera Angle Control, set here to the Low Angle preset. Keeping the angle out of the text means the setting has nothing to argue with.
Example 3: Top-Down. The exact same prompt, run again with Camera Angle Control set to the Top-Down camera type instead. Running one prompt with two camera settings is a quick way to get coverage of a moment: the description stays fixed and only the camera changes. Use it to explore setups for a scene before you rewrite a single word.
{{placeholders}} for the subject and setting.One to three sentences is usually enough. Open with the image type, then the subject, the setting, the light and the framing. Fill each part once with a concrete choice. A short prompt made of precise words gives the model less to misread than a long list of adjectives.
It is a Generation Studio setting for the Nano Banana image models. You pick a preset (Front, Rear, Left Side, Right Side, Low Angle, High Angle or Custom) or a camera type (Top-Down, Wide-Angle or Close-Up), and it steers the prompt. It guides the framing, it does not guarantee it.
Up to 10. Say in the prompt what each one is for, for example the face and costume from one image and the location from another. Giving every reference a clear role makes it easier to keep a character on-model across shots instead of blending everything together.
Yes. Every image lands in the Asset Hub under All Generations. Open the Video tab in the Generation Studio, add the still as the First Frame under Add Keyframes, and write a prompt about the motion, since the picture already decides what is in frame.
Run Nano Banana 2.1 and the other leading models in one AI-native workspace.