daVinci-MagiHuman

What daVinci-MagiHuman is

daVinci-MagiHuman is an open research model: feed a portrait and text or audio, and it generates a lip-synced talking video, with weights under Apache 2.0 for study and reuse.

What you can do with it

• Generate talking videos from portraits
• Study joint audio-video generation methods
• Build derivative tools on the model
• Benchmark against closed alternatives
• Experiment without licensing costs

Who it is for

• Researchers in avatar generation
• Developers evaluating open models
• Teams building on open weights

What to watch out for

• Open weights mean no guardrails: anyone can generate talking videos of any photo, so your deployments carry the consent and disclosure duties
• Generating talking videos of real people without consent is unlawful in many jurisdictions
• Open research models lag commercial products in quality and stability
• Verify the licence and provenance before production use