|
|
The direct reason for the high cost and poor implementation effect is naturally the immature technology. Before large models, the production of virtual humans was mostly driven by real people, which required specialized actors to put on equipment and collect data for a long time through motion capture technology to complete D modeling. This form has high labor costs and a long production cycle. A virtual human practitioner told Self Quadrant that the data collection of a virtual human requires a dedicated production team for several months.
After that, professional technicians are required to perfo Australia WhatsApp Number rm special tuning. This naturally directly increases the production cost of avatars. A certain avatar manufacturer told us: In addition to being driven by real people, there are also virtual people driven by algorithms. However, this type of virtual human requires a large amount of data to train various driver models in the early stage, and the final effect is also affected by speech synthesis, NLP technology, speech recognition, CG rendering and other technologies. Although this type of virtual humans already had relatively stable technology in certain specific directions before large models, they still lacked a powerful brain to unify various modules and allow the virtual humans to achieve the desired effect.
In addition, before large models, virtual humans driven by traditional algorithms usually relied on preset parameters and limited models for training. However, after large models, the generation capabilities of large models gave virtual humans almost unlimited capabilities. training parameters. Apocalypse丨The year of virtual humans Source: AVIC Securities Research Institute Nowadays, large models have penetrated into the entire virtual human production industry chain as a production capacity, which has directly promoted the significant drop in the cost of virtual human production, from the previous level of , to about a thousand yuan.
|
|
發表於 2024-2-20 17:37:18