AI rap video models: original rap vs your own audio, with six real examples
Compare five models through six controlled portrait clips. See what was actually tested, how an original rap differs from an audio reference, and what the samples cannot prove.
Availability note: these are completed controlled tests, not a list of live services. Public generation and purchases are currently closed. The studio and pricing page show current availability.
What we tested on 24 October 2026
Every request used a 10-second, 720p, 9:16 configuration and the orange-studio scene, with no uploaded movement reference. The five solo examples use the same fictional adult portrait; the Mini duo example uses two fictional adult portraits. Original-rap tests use the same short theme. The audio tests use an authorized original track from the first Mini output. Natural body movement is still present. These are single examples of exact combinations, not a broad model ranking.
The exact model and workflow combinations
All files below are actual outputs at 720 × 1280 with an audio track. The requested length was 10 seconds; container durations differ slightly, so the table reports measured file duration.
| Model | Sound route | People | Actual duration |
|---|---|---|---|
| Seedance 2 Mini | Original rap from a theme | One person | 10.080 s |
| Seedance 2 Mini | Own audio | Two portraits | 10.080 s |
| Wan 3.0 | Own audio as a reference | One person | 10.031 s |
| Seedance 2 Fast | Original rap from a theme | One person | 10.080 s |
| Seedance 2.0 | Original rap from a theme | One person | 10.080 s |
| Seedance 2.5 | Original rap from a theme | One person | 10.048 s |
Watch and listen to the six real samples
Press play to load a clip; no video starts automatically. Listen on the same device and volume when comparing delivery and mouth movement. These are our own fictional-adult test assets, not decorative illustrations or borrowed customer work.
Original rap: a theme guides the performance
Mini, Fast, 2.0 and 2.5 were each tested with one person and an original-rap theme. The model creates a short audiovisual performance; a theme is not an exact lyric script. Compare intelligibility, mouth movement, portrait consistency and unwanted visual changes yourself. One successful solo clip does not verify group mode, another language, another duration or 1080p.
Own audio: a reference is not an untouched soundtrack
The Mini duo and Wan solo examples use the same permitted source track. Wan was tested with an audio reference. Its output follows the source closely in this example, but the decoded audio is not byte-for-byte identical and its length differs slightly. Do not assume exact vocals, lyrics, timing or waveform preservation. If an unchanged master track is essential, that requirement needs a separate verified workflow; it is not established by these samples.
Choose a workflow before comparing model names
Start with original rap when you have a theme but no recording. Start with an audio-reference workflow when you already own a short performance and accept that the output may reinterpret it. Then check the live studio for the exact supported people layout, resolution, duration and credit quote. Our planned pricing is informational; it does not enable a model or promise its availability.