Seedream · Seedance · Talking AvatarHyperreal spokesperson pipeline

Talking Avatar (Higgsfield parity)

Upload a portrait and audio clip to generate a lip-synced avatar. FAL “live-portrait” offers instant results; HeyGen requires a project key and publicly reachable assets.

Parity roadmap →Modules overview

1. Configure provider

Provider

FAL Model Slug

Portrait Photo

Audio (wav/mp3)

I confirm I have consent from the person in the photo/audio to generate a talking avatar and store the output.Apply Yoob watermark to generated video

Parity checklist

✅ FAL Live Portrait lip-sync with auto job logging.
🚧 HeyGen proxy endpoint (requires API key + public assets).
🚧 Voice cloning upload flow (Fal fine-tune or external provider).
🚧 Compliance guardrails (consent checkbox, watermark toggle).

2. Preview outputHiggsfield parity

Upload photo & audio to render a talking avatar.

Implementation notes

Fal requests stream via subscribeModel; we enqueue the output in `/api/jobs`.
HeyGen integration requires implementing `/api/heygen/talking` once API access is configured.
Credits check uses `requireCredits('video')`; TODO: show remaining credits inline.
Voice cloning & reference voice upload is scoped for a follow-up sprint.