This guide is for training a face through the API with video URLs. If you prefer a guided flow with upload checks and inline validation, use the PAL Maker.
Outcome
You willPOST /v2/faces with train_video_url, then poll GET /v2/faces/{face_id} until ready. For Phoenix-4.5 preview and tuning status, see Phoenix-4.5 Preview and Status.
Prerequisites
- Publicly downloadable URLs for each asset you send (for example presigned S3 GET URLs). Keep them valid for at least 24 hours after submission.
1. Create the face
Send"model_name": "phoenix-4.5" for Phoenix-4.5. To use Phoenix-4 for full-body framing, include "model_name": "phoenix-4". Phoenix-3 is legacy ("model_name": "phoenix-3") and only for edge cases. See Choose a Model and Training Path.
Personal face (training video):
train_video_url. Cartoon, anime, and Pixar-style humans are fine. Animals, mascots, and other non-human figures are not.
2. Poll face status
Use Get Face:Poll until
status is completed. On Phoenix-4.5 that is the watermarked preview; check finetune_status for the unwatermarked version. See Phoenix-4.5 Preview and Status.3. Start a conversation
Generate a conversation with your face using Create ConversationRelated
- Create Face (full parameter reference)
- Phoenix-4.5 Preview and Status
- Phoenix-4.5 Video Requirements
- Phoenix-4 Video Requirements
- Image to Face if you switch to headshot-based training

