To make an AI music video, choose a musical passage, prepare visual references, describe the action and camera, then generate and review a short clip. This guide walks through a first attempt and shows how to use the clips in a longer music video.
LumiYingUpdated
You need one MP3 or WAV passage, at least one reference image and a visual direction. Start with a 10-second test so that there is one clear musical moment to judge.
Listen to the song and mark a hook, vocal entrance or instrumental phrase with a clear beginning and end. For example, export 00:32–00:42 as a separate 10-second MP3 or WAV. The tool has no audio-trimming timeline, so prepare the excerpt before uploading. Keep it within the selected model's 2–15 or 2–30 second limit and 15 MB file limit.
Start with one clear image showing the performer or scene. Add supporting images only when they answer a specific question: who is the subject, where are they, or what lighting should the shot use? Upload 1–8 JPEG, PNG or WebP images. Avoid conflicting faces, outfits or art styles. In the prompt, identify the role of image 1, image 2 and any other reference.
Describe subject, action, camera and mood in that order. Instead of 'make an epic music video', write 'the singer takes two relaxed steps on the rooftop while the camera moves slowly sideways in warm afternoon light'. Add what should stay consistent. For a first test, use one action and one camera move; add cuts only when you can explain their purpose.
Open AI Music Video Generator and select an available audio-input model. Check its audio limits, upload your music and references, and paste the brief into Visual direction. For the 10-second excerpt, choose a 10-second output. Use 16:9 for a landscape video or 9:16 for a portrait social clip, with references composed for that frame. Output is 720p; review the displayed credit cost before generating.
Generate the clip, then watch its opening, middle and ending. Check whether the subject remains recognizable, the camera move is readable and the final frame is useful. Listen separately for changed vocals, altered musical timing or unwanted sound. Audio guides the model; exact preservation of the original recording and precise cuts on every beat are not guaranteed.
Use Re-edit to restore the original audio, images, prompt and settings. If motion is too busy, remove extra actions or cuts. If the character changes, reduce conflicting references and state the identity details to retain. Keep a useful take before changing the brief. Re-edit prepares the form; generating the next version uses credits again.
Download a useful version. For a full song, plan a few passages and visual beats, generate the required clips and place them in a video editor. Put the original song on the editor's audio track, arrange and trim the clips above it, and make cuts where they serve the phrasing. Use consistent framing and color across clips. This tool generates individual clips; final assembly and lyric titles happen in your editor.
Make these decisions before generating. More references and more cuts are helpful only when they give the model a clearer instruction.
These presets load the reference image, prompt, duration and aspect ratio. Add your own matching audio passage and check the credit estimate before generating.
Use image 1 for the performer, outfit and rooftop setting. Create a 10-second performance shot in warm afternoon light. The performer takes two relaxed steps while the camera tracks slowly sideways, ending on a stable medium shot. Follow the energy of the uploaded music. Preserve the face and clothes; no unrelated dialogue or on-screen text.
Prepare a 10-second hook, load this brief and choose 16:9. This isolates one subject action and one camera move. Review the face and outfit across the whole clip, then listen for changes to your music. If it feels rushed, keep the camera move and remove the steps.
Load this brief
Use image 1 for the two dancers, their outfits and the jazz-club setting. Open on a wide shot, move closer to the dancers, then make one brief cut to their footwork before returning to a clear two-person view. Keep the warm stage lighting and follow the phrasing of the uploaded jazz excerpt. Preserve both dancers; no captions or unrelated dialogue.
Prepare a 15-second phrase and choose 16:9. Start with one wide shot if this version makes too many cuts. Check that both dancers remain distinct and that the footwork detail belongs to the same scene. The prompt guides the cut; playback tells you whether the timing is useful.
Load this briefCompare these public music-video demos for performance, shot changes and atmosphere. Load a reference image and a suggested prompt to try the visual direction with your own music.
A continuous camera move follows a singer onto the stage. Demo: ByteDance. Watch how the continuous camera move connects the dressing room and stage. For a shorter test, choose only one part of that journey.
Follow the singer from the dressing room toward the stage in one smooth tracking shot. Keep her appearance consistent and let movement follow the energy of the uploaded music.
Dance, close-ups and warm stage lighting shape a lively performance. Demo: ByteDance. Watch the change between wide performance shots and footwork details. Give each cut a clear purpose instead of asking for constant scene changes.
Create a vintage jazz-club dance scene. Keep the two dancers recognizable, alternate a wide performance shot with a brief footwork detail, and follow the uploaded music's lively phrasing.
A rooftop setting keeps the performer at the center of the frame. Demo: Freebeat. Watch how the location and performer stay readable during the movement. A simple background makes changes to the subject easier to spot.
Keep the performer on the rooftop. Use relaxed rhythmic movement and a slow lateral camera move, with natural afternoon light and the mood of the uploaded track.
Identify the failure before retrying. Keep the successful inputs and change one cause at a time.
This workflow requires at least one visual reference and a scene description as well as audio. Use a performer image for a performance video or a landscape for an instrumental track, then describe the motion.
Break the song into useful passages, plan the visual beats and generate short clips for them. Assemble those clips against the original song in a video editor. This generator does not automatically divide, arrange or join a whole song.
No. Start with one strong reference. Add images only for a specific subject, location or style detail, up to eight total, and explain their roles in the prompt.
The upload is an audio reference and the output may differ. To preserve the recording in a finished video, use your original song as the final soundtrack in your editor and align the selected clips to it.
Review and download the visuals first, then add timed lyrics and precise cuts in your editor. The generator accepts musical and shot directions but has no lyric-caption editor or beat timeline.
It fills the visual reference and suggested direction. The worked briefs also set duration and aspect ratio. You supply your own music; loading a preset does not start a paid generation.
Create a short AI music video from an audio clip and reference images. Choose an audio-compatible model, describe the scene, and preview your video.
Create a short AI music video from an audio clip and reference images. Choose an audio-compatible model, describe the scene, and preview your video.
Create AI video with audio from text or an image. Describe ambience, dialogue, and sound effects, then generate and review your clip in LumiYing.
Turn a storyboard into a short video. Upload a panel grid or up to eight shot images, describe the motion, and generate one clip with AI.
Keep a good shot going. Upload a video, describe the next action, and use the AI video extender to generate a continuation from your source footage.
You need one MP3 or WAV passage, at least one reference image and a visual direction. Start with a 10-second test so that there is one clear musical moment to judge.
