引导式测验 · 中等 等级
稳定音频潜在扩散测验
使用原始 VAE 和 CLAP 设置、稳定音频 2.0、立体声样本计数和静音填充来研究时序调节的潜在音频扩散。
问题 1 的 8
Stable Audio 使用什么核心技术来生成音频？
预览每个问题8 问题
当您参加测验时，就会出现答案和解释。
- Stable Audio 使用什么核心技术来生成音频？
- 基于规则的 MIDI 排序
- 仅自回归标记预测
- 压缩音频空间中的潜在扩散
- 直接波形记录
- What timing information conditions the original Stable Audio research setup?
- Start position and total-duration signals
- Only a singer’s legal name
- Only a video display resolution
- Only the output filename’s alphabetical position
- 什么组件将原始音频压缩为潜在表示？
- 拼写检查器
- 随机数生成器
- 语音转文本引擎
- 变分自动编码器 (VAE)
- What capability did the official April 2024 Stable Audio 2.0 announcement report?
- A guarantee of six-minute generation in the original 95 sec paper setup
- Up to three-minute 44.1 kHz stereo generation and audio-to-audio transformation
- Only silent mono waveform output
- Only automatic subtitle translation
- How does a text-only generation run in the original latent-diffusion setup start?
- From a noise latent that is denoised using the conditions
- By necessarily recording a live microphone
- By copying a complete training song without a model
- By requiring a complete future waveform as input
- Which text encoder is identified in the original Fast Timing-Conditioned Latent Audio Diffusion architecture?
- Only a handwritten MIDI rule table
- A required raw image-pixel encoder
- An unspecified choice guaranteed identical across all versions
- A CLAP text encoder
- At 44.1 kHz stereo, how many individual channel amplitude values occur in one second before file encoding?
- 44,100
- 88,200
- 22,050
- 176,400
- In the original 95-second window example, a 30-second requested signal is followed by how much silence before trimming?
- 30 秒
- 95 seconds
- 65 seconds
- 125 seconds