Qwen3-Omni Speech Synthesis Process

You are a professional technical architecture illustrator. Please draw a flowchart to demonstrate the **Qwen3-Omni voice synthesis process** based on the following description.

## Overall Requirements

- Purpose: Principle explanation page in a technical sharing PowerPoint.
- Style: Concise, modern **flat 2D flowchart** (not realistic illustrations, not 3D).
- Orientation: Horizontal 16:9 or close to that ratio.
- Text Language: **Simplified Chinese + English abbreviations**, consistent with the text provided below.
- No need for a title, let the meaning be explained by the modules and labels in the diagram.
- **Canvas background should be pure white** (#FFFFFF), avoid using gradients or large color blocks as the background.

## Color Requirements

Please only use the following three main colors as the primary color scheme for modules and visual elements (you can vary the shades):

- `#FCF9E5` —— Basic Token small square for Talker output
- `#ECF8F0` —— Detail Token small square for MTP output
- `#E7F1FE` —— Main color for core modules (such as Talker main network, MTP module, Code2Wav decoder)

Lines and text should be in dark gray or black to ensure readability.

## Layout Structure (from left to right)

Please draw the overall process as a three-stage pipeline from left to right, corresponding to the following ASCII flow:

┌─────────────┐   ┌─────────────┐   ┌─────────────┐  
│ Talker      │ ─▶│ MTP         │ ─▶│ Code2Wav   │  
│ Main Network│   │ Completion Module│   │ Decoder     │  
└─────────────┘   └─────────────┘   └─────────────┘  

### 1. Left Module: Talker Main Network

- Draw a flat rounded rectangle module on the left (main color suggested to use `#E7F1FE`):
  - Module title (two lines):
    - First line: `Talker`
    - Second line: `Main Network`
- On the right side of the **Talker module near the arrow**, draw a row of horizontally arranged small squares, representing basic audio Tokens:
  - The number of squares in this row can be around 8-12, uniform in size and neatly arranged.
  - Fill the squares with color suggested to use `#FCF9E5`.
  - Below this row of squares, label with a smaller font:
    - `Basic Token (content + rhythmic structure)`
- Additionally, below or inside the Talker module, you can add a line of text:
  - `Output: Basic Token (content + rhythmic structure)`

### 2. Middle Module: MTP Completion Module

- On the right side of Talker, connect to the second module with a solid arrow.
- The second module also uses a rounded rectangle (main color suggested to use `#E7F1FE`):
  - Module title (two lines):
    - First line: `MTP`
    - Second line: `Completion Module`
- On the **right side of the MTP module**, based on the row of small squares output by Talker, draw a visual effect of "expanding from one row to multiple rows":
  - The first row should directly continue/align with the row of basic Tokens from Talker (color remains `#FCF9E5`), representing the basic layer utilized by MTP.
  - Above this row of basic squares, add **three rows** of horizontally arranged small squares, creating a total of four stacked layers:
    - The color of the top three rows of squares is suggested to use `#ECF8F0`, representing the detailed audio Tokens generated by MTP.
    - Leave a small vertical gap between each row to clearly show the "stacked" visual hierarchy.
  - Overall, it should look like "three new detail Tokens are stacked on top of the original row of basic Tokens".
- Below this set of multiple rows of squares, label with a smaller font:
  - `Detail Token (audio quality + tone details)`
- You can also add text below or inside the MTP module:
  - `Output: Detail Token (audio quality + tone details)`

### 3. Right Module: Code2Wav Decoder

- Use an arrow to point from the MTP module to the third module.
- The third module is a rounded rectangle (main color suggested to use `#E7F1FE`):
  - Module title (two lines):
    - First line: `Code2Wav`
    - Second line: `Decoder`
- On the **right side of the Code2Wav module**, draw a clear **audio waveform pattern** as the final output:
  - The waveform color can be a slightly darker shade of blue-green, coordinating with the overall color scheme.
  - Below the waveform, label with a smaller font:
    - `Output: Audio waveform (playable audio)`

## Arrow and Connection Relationships

- Use simple, consistent style solid arrows to connect from left to right in the following order:
  - `Talker Main Network` → `MTP Completion Module` → `Code2Wav Decoder` → `Audio waveform`
- Arrow color can be dark gray or slightly darker blue, with uniform thickness.

## Details and Decorations

- The overall canvas background should be pure white (#FFFFFF), without using background gradients.
- You can add some light-colored outlines or subtle shadows in moderation to enhance the module hierarchy, but maintain an overall flat design.
- The overall diagram should allow readers to understand at a glance:
  - Left: Talker generates a row of basic audio Tokens (content + rhythmic structure)
  - Middle: MTP adds three rows of detailed Tokens (audio quality + tone details) on top of the basic Tokens
  - Right: Code2Wav decodes the multiple layers of Tokens into the final playable audio waveform

Please strictly follow the above module names, square layer structure, Chinese instructions, and color requirements for drawing.