Add ElevenLabs voiceover and Playwright CLI browser inspection

[aditya-ig10]

ElevenLabs voiceover:

• Add --no-voiceover and --voice options to invocation
• Check for ELEVENLABS_API_KEY at workflow start; prompt user if missing
• Draft voiceover script from project summary during Step 1, finalize in Step 2
• Generate voiceover audio via ElevenLabs REST API (curl) with fallback to npx hyperframes tts (offline Kokoro-82M model)
• Wire voiceover on track-index 3 with music ducking to 0.12-0.15
• Error handling table for 401/402/429/5xx responses
• Add ElevenLabs API reference to audio.md (endpoint, params, voice IDs)
• Document plan requirement: free tier cannot use library voices via API

Playwright CLI browser inspection:

• Add --url and --no-browser options to invocation
• New playwright-inspect.md reference: open project in real browser, navigate pages, capture screenshots and structured snapshots, extract computed styles (colors/fonts), discover user flows via element interaction
• Screenshots serve as visual references for Hyperframes composition
• Update Step 1 to recommend paired static + browser inspection

[shunithaviv]

Hi @aditya-ig10 ! Thanks for this PR, I went over it and have some thoughts-

Playwright: I think it's a great addition that can help brag choose the right visual style, happy to take this part in.

Voice over: I did some experiments with VO before and honestly I just didn't like the output. In my local implementation I set it to default off and then decided to completely remove the feature. I tend towards a minimal, simple first version as the shipped video, so if we add this back I think it should be:

1. Default no VO unless a user explicitly asks for it. Otherwise the first video a user gets from brag has more layers and probably more bugs, and I'd rather the first experience be simple and reliable.
2. Kokoro as the default. In my implementation I used Kokoro and overall I think it's good enough (+ the bonus of it being already integrated into hyperframes). If a user wants elevenlabs we can include that as an "if a user explicitly asks for elevenlabs" option.

Do you have some samples that you created so I can see how the skill works with VO, with both ElevenLabs and Kokoro?

Would be happy to hear what you think!