> For the complete documentation index, see [llms.txt](https://docs.dreamerland.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.dreamerland.ai/video/minimax-h3/music-video.md). # Music Video The Music Video task turns a song into a video. Upload a track, and MiniMax H3 generates footage that's **synchronized to the music** — cuts, motion and mood that follow the beat, with your track as the soundtrack. Unlike every other H3 task, the audio here isn't generated: it's *your* song, and the video is built to fit it. Because each H3 generation still produces one clip of 4–15 seconds, a full music video is made **piece by piece** — and this task is designed exactly for that: it keeps track of where you are in the song, so every new clip picks up at the right beat. Below example is a full song with a one character reference image. combine with 5 x 15 sec video. {% embed url="" %} ### How It Works 1. In the MiniMax H3 tab, select the **Music Video** task. 2. **Upload the full track** — this is the one required input. Songs up to 5 minutes are supported. 3. Set the **Elapsed video length** (see below). For your first clip, leave it at 0. 4. Upload the **previous video** to continue if it is not your first clip. so the joint between clips is seamless 5. Optionally attach **references** — up to 5 images and a video reference — to keep your performer, band or visual style consistent across every clip. 6. Write a prompt describing the visuals for *this* part of the song, choose the duration (4–15s), and generate. 7. Repeat for the next section of the song, then stitch the clips together in the built-in video editor. ### What Is "Elapsed Video Length"? It's the task's position marker in the song: **how many seconds of the track your earlier clips already cover**. H3 uses it to know which part of the song this clip belongs to, so the visuals sync to the right section — the verse gets verse energy, the drop gets the drop. * **First clip:** elapsed length = 0. The clip opens the song. * **Second clip:** if your first clip was 10 seconds, set elapsed length to **10**. The new clip is generated against seconds 10–20 of the track. * **And so on** — after each generation, add the clip's length to the running total. You type the value rather than slide it, because you read it off what you've already rendered: it's simply the total length of the clips you've made so far. ### The Previous Video Once elapsed length is above 0, a **Previous video** box appears. Attach the clip that came before, and H3 continues it *visually* — same characters, same scene, same lighting — so the joint between clips is seamless. It's optional on purpose: a music video often cuts between unrelated scenes (performance shot → street scene → close-up), and each new scene is a fresh generation that only needs to sync to the song, not to the previous picture. Attach the previous clip when you want **continuity**; leave it out when you want a **cut**. One detail to know: when you attach a previous clip, the new clip inherits its aspect ratio — the ratio control only applies while you're starting fresh. ### Music Video vs. Extend Both tasks make clips that follow earlier clips, so it's fair to ask when to use which. The difference is **what anchors the continuation**: | | **Extend** | **Music Video** | | ---------------- | ------------------------------------------------- | ----------------------------------------------------- | | Anchored to | The previous **clip** — it continues that footage | The **song's timeline** — it continues the track | | Required input | The source video | The full audio track | | Previous clip | Required — it's the whole point | Optional — attach for continuity, skip for a cut | | Audio | Generated by H3, fresh for each clip | Your uploaded song, synced | | Position control | None — it always picks up where the clip ends | Elapsed video length — jump to any point in the song | | Merge option | Yes — can return source + extension as one clip | No — returns the new clip; you assemble in the editor | Rule of thumb: if the **story** is the backbone — one continuous scene unfolding — use **Extend**. If the **music** is the backbone — visuals serving a track, with freedom to cut between scenes — use **Music Video**. (And note that Music Video can do continuity too, via the previous-clip box; Extend just can't do song sync.)