Fix: selecting 0.6B model still downloads and uses 1.7B

[CelebrityPunks]

Summary

• Fixes a bug where selecting Qwen3-TTS 0.6B in the model dropdown still triggers a download of the 1.7B model
• The /generate endpoint was creating the voice prompt before loading the user's requested model size
• Since create_voice_prompt() internally calls load_model_async(None), it fell back to the hardcoded default of "1.7B", causing the 1.7B model to be downloaded even when the user explicitly selected 0.6B
• This reorders the operations so load_model_async(model_size) is called before create_voice_prompt_for_profile(), ensuring the correct model is loaded first

Steps to reproduce

1. Install Voicebox v0.1.12 fresh
2. Download only the Qwen TTS 0.6B model from Model Management
3. Select Qwen3-TTS 0.6B in the generation dropdown
4. Try to generate speech
5. Bug: The app starts downloading the 1.7B model instead of using 0.6B

Test plan

• Select 0.6B model and generate speech — should NOT trigger 1.7B download
• Select 1.7B model and generate speech — should work as before
• Switching between model sizes mid-session should load the correct model

🤖 Generated with Claude Code