AI Video Gen

Keywords → Title → Shorts script → Voicebox audio → MiniMax H3 lip-synced videos (9:16, one per paragraph)

LLM: checking…

1 Keywords → 10 titles

Enter your keywords; the LLM proposes 10 Shorts titles. Edit any title (✏️) or click one to use it.

2 Script (~60 s)

Built from the keywords and your chosen title. Click Edit script to change the text in a popup before continuing.

3 Voicebox narration — one WAV per paragraph

Each paragraph is trimmed to its first sentence before synthesis (change in config.php → audio.first_sentence_only). The TTS model is unloaded from VRAM automatically once all clips are generated (voicebox.unload_after_audio).

4 Persona videos — MiniMax H3 (one per paragraph)

Upload the person's photo — the same face appears in every clip. Rendering runs as a background job — you can close this tab; finished clips appear one by one as they complete (poll: api/video_status.php).

persona preview