gatilin/ming_flash_omni2_audio_dit
0.1B • Updated • 15
Audio Generation Components
Audio generation components, including local DiTs, with native conditioning, history, sampling, and codec dependencies.
Note Ming-flash-omni 2.0 local audio DiT: eight Transformer layers, about 206 MB BF16. Requires native Talker conditioning, audio history and AudioVAE. Preserves CFM/CFG/SDE interfaces; this is not a standalone text-to-speech model.