feat: Odin talking head for Icelandic onboarding

[Peleke]

Summary

• Adds pre-rendered talking head video integration featuring Odin as a guide for Icelandic language learners
• Creates reusable TalkingHeadPlayer component with video controls, autoplay, and error handling
• Integrates Odin videos into 3 onboarding steps: welcome, assessment chat, and congratulations
• Includes documentation for generating videos with ComfyUI + Sonic

Changes

New Components

• TalkingHeadPlayer.tsx - Reusable video player with progress bar, mute, replay controls
• OdinTalkingHead - Pre-configured wrapper for Odin video clips

Modified Files

• OnboardingWelcome.tsx - New "odin-welcome" step with Norse-themed dark UI
• AssessmentChat.tsx - Video panel showing Odin for each AI response (Icelandic only)
• complete/page.tsx - Congratulations video with bilingual transcript

Documentation

• VIDEO_MANIFEST.md - Required video files, scripts, and specifications
• sonic-odin-workflow.md - Step-by-step guide for generating videos

User Flow (Icelandic)

Select Icelandic → Odin Welcome Video → Goals → Assessment Chat (with Odin) → Congratulations Video

Screenshots

N/A - Videos need to be generated using Sonic workflow

Test plan

• Select Icelandic in onboarding - verify Odin welcome screen appears
• Verify video player controls work (play/pause, mute, replay)
• Complete assessment chat - verify Odin video panel appears
• Complete onboarding - verify congratulations video shows
• Test graceful degradation when videos are missing (retry button)

Next Steps

1. Generate Odin portrait image (2.5D semi-realistic)
2. Generate audio with ElevenLabs (Icelandic)
3. Run Sonic workflow to create video clips
4. Deploy videos to public/videos/odin/

🤖 Generated with Claude Code

[Peleke]

🎭 ComfyUI MCP Serverless - Ready for Odin

The ComfyUI MCP serverless deployment is live! Here's what you need for the talking head integration:

Endpoint Info

Item	Value
Endpoint ID	ppvdlc4x2x6deq
Base URL	https://api.runpod.ai/v2/ppvdlc4x2x6deq
Auth	Bearer token (RUNPOD_API_KEY)
PR	Peleke/comfyui-mcp#9

API Calls

1. Health Check:

POST /runsync
{"input": {"action": "health"}}

2. Portrait Generation:

POST /runsync
{"input": {
  "action": "portrait",
  "description": "Odin, Norse god, one eye, long gray beard, wise ancient appearance, 2.5D semi-realistic style",
  "width": 768, "height": 1024, "steps": 20
}}

3. TTS (Voice Cloning):

POST /runsync
{"input": {
  "action": "tts",
  "text": "Velkomin, traveler...",
  "voice_reference": "<path_or_url_to_voice_sample>"
}}

4. Lip-sync (async):

POST /run
{"input": {
  "action": "lipsync",
  "portrait_image": "<base64_or_url>",
  "audio": "<base64_or_url>"
}}
# Poll with GET /status/{job_id}

Current State

• ✅ Endpoint created and healthy
• ⚠️ Workers scaled to 0 (no charges) - wake before use
• ⚠️ Models not yet loaded (need to add SDXL/Flux, F5-TTS, SONIC)

To Generate Odin Videos

# 1. Wake workers
gh workflow run gpu-control.yml -f action=wake -R Peleke/comfyui-mcp

# 2. Generate portrait, TTS, lip-sync via API

# 3. Download videos to interlinear/public/videos/odin/

# 4. Sleep workers (save $$$)
gh workflow run gpu-control.yml -f action=sleep -R Peleke/comfyui-mcp

TypeScript Client

Available at comfyui-mcp/src/runpod-serverless-client.ts:

const client = new RunPodServerlessClient({ endpointId, apiKey });
await client.portrait({ description: "Odin..." });
await client.tts({ text: "...", voice_reference: "..." });
await client.lipsync({ portrait_image: "...", audio: "..." });

---

From comfyui-mcp PR #9