A useful FLUX 3 music video prompt describes a shot with a job to do: establish a mood, build energy, reveal a new space, or give the viewer a pause. Decide where the shot belongs in your song, then describe what happens in the frame and how the camera should observe it.
Do not start by asking one paragraph to explain an entire music video. A short, coherent visual idea is easier to review and revise. The examples below are original writing exercises, not FLUX output samples or evidence of a paid generation test.
Use the video model and choose the starting material
Black Forest Labs distinguishes FLUX 3 video generation from the FLUX.2 and FLUX.1 image families. Check the model selector before applying a guide written for still images to a moving scene. BFL model overview.
The documented FLUX 3 routes start from a text prompt, pinned image frames, or an existing clip to continue. FLUX Video Edit is a separate tool for changing an existing clip. FLUX 3 video documentation.
| What you already have | A sensible starting route | What the prompt should explain |
|---|---|---|
| A visual idea with no chosen frame | Text-to-video | The scene, action, and camera behavior |
| An approved opening image | Image-to-video | What moves from that starting composition |
| A planned opening and ending | Keyframes, where supported by your interface | The physical transition between those states |
| A clip that needs to continue | Video continuation | What happens next, including the camera's existing motion |
Choose the route before writing. Describing a new room is a different task from extending an already established camera move.
Give each shot a musical purpose and a time window
Listen to the selected recording and mark a few useful moments: the opening vocal, the arrival of a chorus, a quieter bridge, and the final phrase. Use the actual recording rather than assuming repeated sections are identical.
Then make a small shot note:
Section: first chorus
Purpose: make the space feel larger
Start: chosen timestamp in the final recording
Length: 8 seconds
Opening: a close view of one paper crane
Ending: a wider room filled with hanging paper cranes
Camera: one slow pullback
Soundtrack: keep the existing song in the final edit
The eight-second length is illustrative. At a steady 120 BPM in 4/4, four bars occupy eight seconds: four bars multiplied by four beats, with half a second per beat. A pickup, pause, or tempo change can alter the useful edit points, so listen before setting the cut.
BFL's Playground lets you choose a duration between five and twenty seconds for its video-generation workflow and set the aspect ratio, or leave them on auto. For a planned edit, choose the available controls deliberately and inspect the returned clip. Do not rely on a duration written only in the prompt. Playground guide.
Describe a visible change instead of a list of adjectives
“Emotional, beautiful, epic” may express your intention, but it leaves the actual event undecided. Replace part of that description with something a viewer could point to.
| Vague direction | A more observable instruction |
|---|---|
| The scene becomes hopeful | The curtain lifts and daylight reaches the back of the room |
| Make the chorus bigger | Pull back from the small object to reveal the larger surrounding space |
| Add cinematic movement | Track slowly sideways while keeping the subject at the same apparent size |
| Make the ending peaceful | Let the last movement settle and hold the composition |
A practical first prompt contains a subject, a setting, one principal action, a camera move, and an ending condition. This is an editorial checklist, not a required FLUX syntax. Use ordinary sentences when they communicate the shot clearly.
Separate subject motion from camera motion. “The crane turns as the camera circles it” asks for two changes. “The crane remains still while the camera moves sideways” gives you a simpler result to evaluate.
Three original FLUX 3 prompt examples for a music video
These examples share a paper-crane motif so the sequence has a visual connection. Change the motif to fit your own song. Set the duration and frame shape using the controls available in the product you use.
A restrained verse: one object, one small movement
A small ivory paper crane rests on a dark upright piano in a quiet
rehearsal room. Warm afternoon light falls across the keys. A gentle
draft lifts one folded wing slightly, then it settles. The camera
tracks slowly to the right at the height of the piano keys, keeping
the crane clearly separated from the background. Finish with the
crane still in the same patch of light.
This gives the verse a contained action. When reviewing it, check whether the object keeps its shape and whether the camera movement remains calm. If the model adds a busy event in the background, clarify the room and simplify the action before adding more stylistic language.
A chorus reveal: widen the space without changing the motif
Begin close to an ivory paper crane hanging from a thin thread in
a large rehearsal hall. The camera slowly pulls backward to reveal
dozen after dozen of matching paper cranes suspended above the floor.
They sway gently in the same light breeze. Tall windows illuminate
the room with soft daylight. Keep the pullback continuous and finish
on a wide view of the hanging arrangement.
The reveal supplies the change in scale. You do not need several unrelated locations to make the chorus feel different from the verse. Review the opening and final compositions together: does the wider image deliver the contrast you intended?
A bridge or ending: let motion resolve
One ivory paper crane floats in a shallow, still pool inside an empty
courtyard. Small ripples carry it slowly toward a patch of reflected
sky. The camera remains fixed just above the waterline. The ripples
become quieter as the crane reaches the reflection. End on the settled
water and the crane holding its position in the frame.
This is a direction to test, not a guarantee of physically perfect paper on water. If the movement or material behaves implausibly, shorten the event or choose a less ambiguous setup. Keep the element that matters to the song rather than defending every detail of the original prompt.
Use reference frames to reduce decisions, not to remove review
An approved image can settle composition, palette, clothing, or a recurring object before you ask for movement. BFL documents image-to-video keyframes as frames within the clip, including opening and ending images. Video-generation capabilities.
For the chorus example, prepare compatible close and wide views of the same hanging arrangement. If the close-up contains a red crane in a dark room but the ending shows white cranes outdoors, the model has to solve a change you may not actually want.
Check these pairs before generating:
- Does the recurring subject have the same defining appearance?
- Are the room, light direction, and palette compatible?
- Can the requested camera move plausibly connect the compositions?
- Is there enough space around the subject for the final video crop?
A reference image does not replace a motion instruction. Say what should move and what should remain visually stable, then inspect continuity in the result.
Keep the original song separate from generated audio
FLUX 3 can generate sound along with its pictures. BFL documents generated audio as enabled by default. That is useful for some scenes, but it is not the same as synchronizing a supplied Suno recording. Audio and defaults.
If your song is already finished, decide whether you need any generated sound. Disable it where your chosen interface supports that control, or mute the clip's audio in your editor before adding the original recording. Keep any desired effects on a separate track so they do not accidentally replace or double the music.
Likewise, a prompt that describes a singer performing does not prove the lips match your existing vocal. If mouth timing is essential, evaluate an appropriate audio-driven performance workflow separately. If the words mainly need to be readable, a lyric-led scene may be enough.
Change the smallest part that failed
Review a draft against the shot note, not against a general impression that it should be more cinematic.
| What is wrong | What to change first |
|---|---|
| The camera changes direction unexpectedly | Replace competing moves with one direction and an ending composition |
| The frame becomes too busy for lyrics | Simplify background action and reserve a quieter area for text |
| The recurring object changes appearance | Reuse an approved frame and keep its identifying description consistent |
| The reveal happens before the musical moment | Adjust the shot's place in the edit or redesign when its main action occurs |
| The clip introduces an unwanted voice or score | Review audio settings and the clip's audio track in the edit |
BFL also documents draft generation and a separate enhancement path for an accepted draft. A fresh full-quality generation is not necessarily the same operation as enhancing the draft you chose. Use the documented path your interface provides and review the final file. Draft mode.
Avoid changing the motif, camera, lighting, and timing together unless you intend to abandon the shot. One focused revision makes it easier to see whether you solved the actual problem.
Assemble the visuals around the finished song
Place the accepted clips on a timeline with the original recording. Check the relationship between the cut and the musical event, then review each join for continuity. A strong isolated clip can still be the wrong length or energy for its location in the edit.
Add readable lyrics after the main visual rhythm is settled. Keep faces, focal objects, and important motion clear of the text. Review the phone-sized composition as well as the larger preview.
If your main goal is a lyric video from existing audio, LyricVideoMaker's audio-to-lyric-video workflow is an alternative starting point. The AI music video page and lyric video guide explain related production choices. This article's FLUX examples describe use of a separate video-generation tool, not a newly announced FLUX integration in LyricVideoMaker.
Can one prompt generate an entire music video?
The official release describes short clips, with up to twenty seconds in one generation. Plan the song as a sequence of shots, and check the limits for the specific mode you select. A whole-song concept is useful for planning, but each generated clip still needs a place in the edit.
Do I need exact timestamps inside every prompt?
Use timestamps in your own shot plan to make placement unambiguous. Where a tool supports timed keyframes, follow its documented format. Text describing a musical moment should not be treated as proof that the returned movement lands on that exact beat.
Are these tested FLUX video presets?
No. They are original examples of clear shot descriptions. Judge your generated clip against your song and its visual purpose, and keep only the prompts and results that help you complete that edit.
