Fix AssemblyAI transcript model

[FujiwaraChoki]

Summary

• replace the deprecated AssemblyAI nano transcript model default with the supported universal model
• normalize legacy/unsupported transcript model env values to SDK-supported AssemblyAI enums
• update Docker Compose and .env.example defaults so backend and worker use the supported model

Root Cause

AssemblyAI now rejects nano during transcription, which caused local API video generation to fail at the transcript step.

Validation

• python3 -m py_compile backend/src/config.py backend/src/video_utils.py
• Ran local Docker backend/worker and generated clips through the signed API for https://www.youtube.com/watch?v=JiEFDJxGx2s
• Completed task bc3c82d8-bc4c-4490-9821-7cc7788d6c1d with 4 clips
• Downloaded and ffprobe-checked clip 1: H.264/AAC, 1080x1920, 30fps, 40.301s

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
supoclip	Ready	Preview, Comment	Jun 28, 2026 10:34pm