Guide
Visuals that move with your music
A music video is a sequence of moods. Render each one as a short clip, cut them to the beat, and let the tiers decide how polished each moment needs to be.
Sample renders
More examplesThree kinds of shot
Most music videos are built from performance shots, narrative or mood shots, and abstract inserts. Performance shots show someone playing or singing under stage light and are the most demanding: hands, instruments and faces move fast and need the Cinematic tier to stay coherent. Mood shots, such as a car at night, a rooftop at dusk or a rain-streaked window, hold up well on Standard. Abstract inserts, such as ink in water, light leaks or particle fields, look good even on Draft and are cheap filler between the expensive moments.
Budget accordingly. A typical three-minute video might need six performance shots, ten mood shots and a dozen inserts. Spend Cinematic credits on the performance shots that will be on screen for the chorus and let the rest run on cheaper tiers.
Lighting language for stage scenes
Stage lighting is where the models shine, provided you name it. "Backlit with haze", "two-colour wash in teal and magenta", "single hard spotlight from above", "strobe pulses" and "warm tungsten practicals" each produce a distinct look. Combine one key description with one atmosphere word such as haze, smoke or fog, and the renders gain the depth that flat lighting lacks.
For a consistent artist across shots, describe them the same way every time: hair, outfit colour, instrument. Or animate from a photo of the performer, which keeps the likeness faithful in the first frame of each clip.
Sound on or sound off
Your track is the soundtrack, so for most shots you will switch generated sound off and lay your own audio over the cut. The exception is performance footage: leaving sound on produces a plausible live-room ambience you can mix quietly under the master, which helps sell the shot as real. The picture renders identically either way.
- Sound off for mood and abstract inserts
- Sound on, mixed low, for performance shots
- Match clip length to a bar count at your tempo
- Use 9:16 for vertical platforms, 16:9 for the main cut
Cutting to the beat
Work out how many seconds a bar lasts at your tempo and render clips in multiples of that. At 120 beats per minute a bar is two seconds, so four-bar clips are eight seconds, which both Standard and Cinematic support. Rendering to the bar makes the edit almost assemble itself.
Vertical versions are a separate render rather than a crop. Set the aspect ratio to 9:16, reuse the same prompts, and the model recomposes each shot for the tall frame instead of leaving you with cut-off heads.
Tier specifications
| Tier | DraftFast previews | StandardEveryday quality | CinematicTop fidelity |
|---|---|---|---|
| Model class | Lightweight diffusion video model | Compact audio-video model | Full-size audio-video model |
| Resolutions | 720p / 1080p | 480p / 720p | 480p / 720p / 1080p |
| Clip length | 1–20 s | 4–15 s | 4–15 s |
| Frame rate | 24 fps or 48 fps | 24 fps | 24 fps |
| Sound | Optional track | Generated with the picture | Generated with the picture |
| Credits per second at 480p | — | 3 | 6 |
| Credits per second at 720p | 1 | 6 | 12 |
| Credits per second at 1080p | 3 | — | 30 |
Questions
- Can the model generate the music?
- No. Standard and Cinematic generate ambient sound and effects that match the picture, not songs. Bring your own track and mix it in your editor.
- Will the performer look the same in every clip?
- Consistency improves when every prompt describes the performer identically, and improves further when you animate from a photo. Expect some drift between clips, which cuts hide well.
- Which tier for a chorus shot?
- Cinematic at 720p. Fast hand and face movement under coloured light is exactly what the top tier handles better than the others.
- How do I make a vertical and a wide version?
- Render each prompt twice with different aspect ratios. Compare mode is handy here too: queue 16:9 Standard against 9:16 Standard on the same prompt and get both at once.
- Who owns the video clips I make for my song?
- You own the clips, per /terms-of-service, which assigns you all output produced at your direction. Copyright law for generated footage is not the same everywhere, so consult a lawyer for your territory if you plan a commercial release.
Render the chorus shot
Describe the performer, the light and one camera move. Draft it, then promote the one that works to Cinematic.