Dance reconstruction · DA3 → character reference → H3 5 reconstruction comparisons using 12-second clean source excerpts, with matching depth and source audio. Each reference column explains which appearance information is supplied. Generated outputs are separate from the clean corpus.
Videos preload automatically. Play all comparisons together, or use Play comparison on one row. One selected audio track plays at a time. Review identity, choreography, timing and artifacts; decisions stay in this browser until exported.
Earlier six-second pilot
Play all Pause all Restart all Audio Generated · first comparison Source · first comparison Muted
Export review JSON 1. NEXZ · Saucin' like this
Source · 12 seconds Original audio supplied to H3 DA3 depth Depth reference; white is nearer H3 result 768 × 1344 · 12.25s · generated audio · unreviewed Nano Banana reference Same dancer, hair, outfit and proportions requested
Play comparison Pause all Restart all Audio Generated Source Muted
Decision Unreviewed Good reconstruction Identity drift Motion mismatch Audio/timing mismatch Artifacts
2. 1MILLION Dance Studio · Shout-out to @__zpdlzk 👀✨ FLOWY BOM-BOM 🌊 @harimuuu #choreography · group-preservation fix
Source · 12 seconds Original audio supplied to H3 DA3 depth Depth reference; white is nearer H3 result 768 × 1344 · 12.25s · generated audio · unreviewed Full-scene reference All visible people retained; the original source frame also supplies scene layout.
Play comparison Pause all Restart all Audio Generated Source Muted
Decision Unreviewed Good reconstruction Identity drift Motion mismatch Audio/timing mismatch Artifacts
3. aespa · Watch me go go - Winter
Source · 12 seconds Original audio supplied to H3 DA3 depth Depth reference; white is nearer H3 result 768 × 1344 · 12.25s · generated audio · unreviewed Nano Banana reference Same dancer, hair, outfit and proportions requested
Play comparison Pause all Restart all Audio Generated Source Muted
Decision Unreviewed Good reconstruction Identity drift Motion mismatch Audio/timing mismatch Artifacts
4. aespa · Photo identity · original dance outfit
Source · 12 seconds Original audio supplied to H3 DA3 depth Depth reference; white is nearer H3 result 768 × 1344 · 12.25s · generated audio · unreviewed
Play comparison Pause all Restart all Audio Generated Source Muted
Decision Unreviewed Good reconstruction Identity drift Motion mismatch Audio/timing mismatch Artifacts
5. aespa · Photo appearance · black top
Source · 12 seconds Original audio supplied to H3 DA3 depth Depth reference; white is nearer H3 result 768 × 1344 · 12.25s · generated audio · unreviewed Supplied photo · identity + visible clothing Supplied Wikimedia photo Photo: K-POPIT 케이팝잇 / Wikimedia Commons · CC BY 4.0 Reference-only R2V input; no fixed first frame.
Play comparison Pause all Restart all Audio Generated Source Muted
Decision Unreviewed Good reconstruction Identity drift Motion mismatch Audio/timing mismatch Artifacts
Same dancer, hair, outfit and proportions requested
All visible people retained; the original source frame also supplies scene layout.
Same dancer, hair, outfit and proportions requested
Additional outfit-only referenceSupplied Wikimedia photoPhoto: K-POPIT 케이팝잇 / Wikimedia Commons · CC BY 4.0Reference-only R2V input; no fixed first frame.
Supplied Wikimedia photoPhoto: K-POPIT 케이팝잇 / Wikimedia Commons · CC BY 4.0Reference-only R2V input; no fixed first frame.