The Formula
[camera] shot of [subject] [action] in [environment]. [supporting visual and motion details]
- Camera
- One framing term and one movement term, such as a low tracking shot.
- Subject
- Who or what the shot is about, with a fixed description you can repeat across shots.
- Action
- One clear thing the subject does.
- Environment
- Where it happens: place, time of day, weather.
- Supporting details
- Light, palette, motion quality, and the sound: speech, ambience, effects, music.
The Rules
Direct the scene
Black Forest Labs puts it plainly: think of the prompt as directing a scene, not describing a collection of objects. Say what happens, how subjects move and how the camera behaves.
One framing, one move, one action
"Low tracking shot" is clear; "low aerial handheld orbit push-in" is not. Give the subject one clear action.
Start short, then lock details
A short prompt lets the model fill in framing and mood; a longer one gives you control. Over-stuffing makes motion less coherent, so add detail only where it improves control.
Quote the words and name who says them
A visible speaker gives FLUX 3 a face to lip-sync. Off-screen lines need a "voiceover" or "narration" cue, or the quote may render as text in the frame.
Name sound sources, never silence
"Rain ticking against a metal awning" beats "moody ambience". Instead of asking for silence, name a quiet source such as rain on the window.
Direct the voice with anchors
Give a person, register, recording setup and delivery instead of stacks of adjectives, and add guardrails like "no announcer delivery" when needed.
Repeat the subject description across shots
Use the same fixed subject description, word for word, in every shot's prompt to keep the character consistent.
Use timesteps when beats must land
For action that has to hit marks, write a short timeline with two or three beats per 5 seconds, and mark a hard cut where the angle changes.
A Template You Can Copy
[Framing and one movement] shot of [subject] [one clear action] in [environment]. [Supporting visual and motion details: light, palette, motion quality]. [Speaker] says, "[exact words]". [Delivery: person, register, recording, tone]. Ambience: [the place]. Effects: [sounds tied to visible actions]. Music: [style and where it sits in the mix]. No on-screen text, no subtitles.
A Real Example
The shot: Night, rain-streaked diner window. A woman in a red coat sits alone in the last booth, turning a ring on the table. Slow push in from the window. Mara says: "You kept it." (6 seconds, 16:9)
Slow push-in shot of a woman in a red coat sitting alone in the last booth of a rain-streaked diner at night, slowly turning a ring between her fingers on the table. Warm interior light glows against the cold, rain-streaked window, her eyes lowered in quiet thought as the camera moves closer from outside the glass. Rain ticks softly against the window, and the low hum of an empty diner fills the background. She lifts her eyes and says, "You kept it." (no on-screen text, no subtitles)
It opens in FLUX 3's one-liner shape with a single movement term (slow push-in), then adds the light and her one clear action. The sound is named by source, the line is quoted with a visible speaker so it can be lip-synced, and "no on-screen text, no subtitles" stops the quote from being printed in the frame.
Dialogue
Put the exact words in quotation marks and say who delivers them; a visible speaker is lip-synced. Mark off-screen lines as voiceover or narration, and add "no on-screen text, no subtitles" when the line must not appear as text. Keep lines short and sayable, and leave room in the clip for the line to finish.
The mechanic says, "Try it now." Quiet, matter-of-fact delivery.
Camera Moves
| Move | How to write it for FLUX 3 |
|---|---|
| Tracking | A low tracking shot of a fox sprinting through wet pine undergrowth at dawn. |
| Slow push in | Slow push-in shot of a woman sitting alone in the last booth of a diner. |
| Handheld follow | Handheld follow shot of a penguin waddling across sun-warmed boulders. |
| Overhead drift | Overhead drift shot of a market stall as the vendor stacks oranges. |
| Static | Static shot of a still harbor at dawn, boats motionless on glassy water. |
| POV | POV shot of a boxer weaving through a dim training gym. |
Common Mistakes
| Mistake | Fix |
|---|---|
| Stacking camera terms like "low aerial handheld orbit push-in". | Use one framing term and one movement term. |
| A quoted line with no speaker or voiceover cue. | Name a visible speaker or say voiceover, and add "no on-screen text, no subtitles". |
| Asking for silence or "moody ambience". | Name concrete sources such as rain against the window or a refrigerator hum. |
| Voice adjectives like "professional, warm and engaging". | Give a person, register, recording setup and delivery, plus "no announcer delivery" if it sounds like an ad. |
| A long line in a short clip, so the last word is cut off. | Shorten the line or raise the duration. |
Length, Resolution and Price
FLUX 3 makes 5 to 20 second clips at $0.17 per second at 720p or $0.29 at 1080p, with optional sound, lip-synced speech in 13 languages and an end frame, in 21:9, 2:1, 16:9, 4:3, 1:1, 3:4 or 9:16. Its keyframe mode pins up to 10 stills on the timeline. FLUX 3 Draft is a 720p preview tier at $0.06 per second for testing motion before the full render (checked 2026-10-22). Every model side by side is on the model spec sheet.