AI disclosure
Effective date: 28 October 2026 · Last updated: 28 October 2026.
Tiny Desk Video AI is operated by wen yong, an individual based in China, at tinydeskvideoai.com. Contact: support@tinydeskvideoai.com.
Tiny Desk Video AI integrates MiniMax-H3-Max through MiniMax's international API. The maker supports a 10-second, 9:16 office-band parody at 768P, directed as one generated clip with four short scene beats; exact camera cuts and timing can vary. One authorized adult portrait is used for each assigned role: lead and up to four optional companion musicians. Unassigned musicians are fictional. Uploading role photos and checking a cast do not consume generation credits or generate a video. New band submission is available only while the updated processing service is running and all submission checks pass. Audio is MiniMax-native original funk accompaniment with wordless vocals. The service does not include or license the original Outstanding recording. Previously submitted 10-second tasks retain their frozen model, audio and delivery settings.
The output depicts a generated performance, not evidence of an actual performance or endorsement. Facial likeness, body and finger anatomy, instruments, musical timing and sound can contain errors. Exact identity preservation, accurate fingering, beat-by-beat synchronization and singing lip synchronization are not guaranteed. Outputs are not guaranteed to be unique or legally protected.
The service is independent of MiniMax, NPR, Tiny Desk Concerts and any artist. Do not represent an output as documentary footage, an authentic statement or an endorsement. Follow applicable synthetic-media disclosure rules when sharing it.
Use a portrait only with the necessary permission and follow applicable law, the Acceptable use policy and provider restrictions. An AI label does not establish consent or clear rights to an image, likeness, music or other material.
Instruction checks and delivery
Waffo PromptSift scans the generation instructions. This integration checks text; it does not scan the uploaded photo, generated video frames or generated audio track. The model provider may refuse a request under its own safety controls. A permitted instruction and a technically completed file do not certify the content's safety or establish consent, identity or copyright permissions.
When generation succeeds and technical delivery checks complete, the result is made available privately to the account owner. A pending, blocked or unavailable instruction check prevents normal submission. If the provider outcome is uncertain, keep the original task reference and do not create a replacement paid submission; contact support for help.
A completed file being technically playable does not certify its truth, permissions or safety. Review the result before sharing. There is no claim that every download already contains a permanent AI watermark; your own publication must still disclose synthetic media as required.
See Privacy for photo processing and Acceptable use for restrictions and reports.