[SageMaker] New Qwen-TTS generation example

[ehcalabres]

Description

This PR adds a new SageMaker example about speech generation with the Hugging Face vLLM Omni DLC and the Qwen/Qwen3-TTS-12Hz-1.7B-Base and Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice models. The topic of the example is: Deploy Qwen3-TTS for ad voiceovers on Amazon SageMaker AI.

This PR needs #2649 to be merged first, as that PR contains improvements for SageMaker notebook docs.

---

Note

Low Risk
Documentation-only addition (notebook and images); no changes to application or infrastructure code in the repo.

Overview
Adds a new SageMaker SDK notebook under tts-qwen-sagemaker-vllm-omni titled Deploy Qwen3-TTS for ad voiceovers on Amazon SageMaker AI, plus cover and UI screenshots in assets/.

The walkthrough deploys Qwen3-TTS CustomVoice on the Hugging Face vLLM-Omni DLC (ml.g5.xlarge) via SageMaker Python SDK v3 Model / EndpointConfig / Endpoint, with HF_TASK=text-to-speech and invocations routed through CustomAttributes="route=/v1/audio/speech". It also documents an optional Base endpoint for authorized voice cloning, a Gradio Ad Voiceover Studio that calls those endpoints, and optional MCP exposure via gradio[mcp].

^{Reviewed by Cursor Bugbot for commit 6791327. Bugbot is set up for automated code reviews on this repo. Configure here.}

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[dwarez]

thank @ehcalabres, very cool 😄

I ran the notebook end to end and I confirm it's working, however the cloned output is a bit ... weird? 😅 Did you get a nice output from that during your tests? I used the custom-voice-response.wav previously generated audio as reference input.

[cursor]

Cursor Bugbot has reviewed your changes using default effort and found 1 potential issue.

There are 3 total unresolved issues (including 2 from previous reviews).

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 9133628. Configure here.}

[ehcalabres]

Hey @dwarez! So, I tried a couple of times initially and it worked good with my own voice. I tried both with and without the transcript (using x_vector_only_mode through the gradio app). Having said so, I tried later with a different audio and it actually returned nonsense, so I'm not sure how reliable it is. Let me try again with a couple of samples. Maybe we can add a disclaimer for that