Shortform music videos are 7–60 second vertical clips built around a strong audio hook. Success depends on immediate visual identity, beat-aligned edits, and clear on-screen messaging.

Key specs:

• Aspect: 9:16 (1080×1920 or 2160×3840)
• Durations: 7–15s (hook-first), 20–35s (story + drop), 40–60s (full chorus)
• Priorities: 1) Hook in first 1.5s, 2) Character/brand recognizability, 3) Beat-synced motion and text

• Pre-production: Ingest character bible; define concept, emotional beat, and CTA.
• Script + beat map: Detect BPM, mark hook, map cuts/moves per bar.
• Asset prep: Keyframes, control refs, character poses, lyric text.
• Generation: Image/video synth, control passes, compositing.
• Captions/graphics: Auto-lyrics, kinetic type, branding.
• Mastering/export: Loudness, codec, aspect, thumbnails.
• Publish/iterate: A/B hooks, watch-through analysis, remix variants.

Lock visual identity before you cut to music. Feed your model or pipeline a concise bible so characters remain consistent across shots and episodes.

What to include:

• Identity: Name, age, archetype, 1‑line backstory
• Visual: Hair, eyes, wardrobe, palette, accessories, silhouette
• Style rules: Line weight, shading, grain, color grade, lens look
• Motion tags: Walk cycle, idle loop, combat poses, dance identifiers
• Do/Don’t: Canon limits (no glasses off; jacket always zipped, etc.)

Implementation notes:

• Provide 6–12 curated reference images with consistent lighting and angles.
• Tag references with pose/shot labels (CU, MCU, WS, profile, 3/4).
• Use a short, reusable style string in prompts; keep it stable across scenes.

Treat music as the timeline. Map story beats to bars before generating visuals.

Process:

1. Detect BPM and structure (intro, hook, drop, outro).
2. Mark timecodes for: first impact (≤1.5s), hook entry, drop, loop point.
3. Create a shot list indexed by bar/beat (e.g., 00:00.00, 00:01.00, etc.).
4. Assign camera moves, transitions, and text events to specific beats.

Useful math:

• Frames per beat = (FPS × 60) / BPM
• For BPM 120 at 30 FPS: 15 frames/beat; place cuts at multiples of 15.

• Anime AMV loops: Character hero pose → impact frame → speed ramp.
• Manga panel slides: Panel wipes on off-beats; halftone overlays.
• Kinetic lyric type: Phrase-per-beat pop-ins, outline + drop shadow.
• Stylized rotoscope: Motion-consistent edges; glow on downbeats.
• VTuber/Avatar lip‑sync: Viseme-driven mouth; camera punch-ins on chorus.

• Use viseme mapping or auto lip‑sync; bake at 24–30 FPS.
• Offset mouth shapes slightly ahead of consonant hits (1–2 frames) for perceived sync.
• Key on phrasing, not every phoneme; emphasize downbeats and rhyme ends.
• If looping, align mouth to a neutral closed pose at the loop point.

• Match cuts on motion vectors (whip → whip).
• Zooms + dolly-ins for pre-chorus tension; snap-out on drop.
• Strobe/glitch only on isolated beats to avoid fatigue.
• Light leaks, chromatic aberration, and halftone pulses bound to kick/snare.

• Auto-generate lyric timing, then hand-tweak key lines for emphasis.
• Accessibility: High-contrast text, 2–4 px stroke, drop shadow, safe area padding.
• Place CTA microcopy (e.g., “Save sound”) after chorus; never block faces.
• Keep reading level simple; 6–8 words per burst.

• Resolution: 1080×1920 (preferred), H.264/HEVC; high profile, 10–20 Mbps.
• Audio: -14 to -16 LUFS integrated; true peak ≤ -1 dBTP; 48 kHz.
• Color: sRGB/Rec.709; avoid crushed blacks; mild sharpening only.
• Provide multiple loop points; ensure seamless audio crossfade if needed.

Use licensed or platform-cleared audio. If using trending sounds in-app, mute your upload’s music track and align visuals to the platform sound. Keep documentation for commercial tracks and distribute alt mixes for multi-platform posts.

Beat‑mapped shot list (copy and adapt):

• 0.0s (Hook sting): CU character snap to camera; text: “I can’t lose this time.”
• 1.0s: Whip right → WS; jacket flutter; lyric word 1 pops.
• 2.0s (Downbeat): Impact frame; glow pulse; logo 0.3s.
• 3.0s: Zoom to MCU; lip‑sync phrase; speed ramp 0.75x→1.25x.
• 7.0s: Loop point on neutral pose.

Character‑driven MV prompt core:

• “Anime hero, teal/orange night city, medium shot, jacket stripes, silver pin, sharp ink lines, soft bloom, clean halftone shadows, 35mm lens look, steady cam 2%, subtle film grain.”
• Keep this style string identical across shots; only swap pose/camera.

• Hook hold (first 3s): target ≥70%.
• Average watch time: optimize to full duration; enable loops.
• Saves/shares: add utility (lyrics, trend, tutorial) or strong identity.
• A/B test: first frame, caption wording, thumbnail, and loop point.

• Off‑beat cuts: Recompute frames/beat; snap edits to grid.
• Character drift: Tighten bible; reduce prompt randomness; reuse seeds.
• Muddy captions: Increase stroke; restrict background motion under text.
• Overlong intros: Front‑load identity and motion in ≤1.5 seconds.

• Beat detection: Audacity, Ableton Live (warp), Adobe Audition markers.
• AI video: Runway Gen‑3, Pika, Stable Video (AnimateDiff), Deforum workflows.
• Control: OpenPose/Depth/Lineart control nets for motion consistency.
• Edit/composite: CapCut, Premiere Pro, After Effects (Time Remap + Markers).
• Captions: CapCut auto-captions, Premiere Speech to Text.