Skip to content

Latest commit

 

History

History

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
 
 
 
 
 
 

README.md

Gemini Omni Flash video — Google Flow API batch generation (Node.js)

Batch-generate Gemini Omni Flash audio-native video through the Google Flow API by useapi.net — synced dialogue, reference images, and video-to-video edits from a list of prompts.

📖 Full walkthrough: Generate audio-native AI video with Gemini Omni Flash

omni-flash.mjs reads prompts from prompts.json, uploads any reference images or source videos, submits each job to POST /videos in async mode with model: "omni-flash", polls GET /jobs/{jobId}, and downloads every finished MP4.

Prerequisites

  • Node.js v21 or newer (no dependencies to install — uses built-in fetch)
  • A useapi.net API token
  • A connected Google Flow account email (Omni Flash runs on Plus, Pro, and Ultra plans)

Usage

node ./omni-flash.mjs <API_TOKEN> <EMAIL> [PROMPTS_FILE]

PROMPTS_FILE defaults to prompts.json. The script looks the account up by email and checks its health field before submitting.

Prompts

prompts.json is an array of prompt objects — prompt is the only required field; everything else falls back to the API defaults (model omni-flash, landscape, 8 seconds; durations 4 / 6 / 8 / 10 s).

  • Spoken dialogue: set referenceAudio_1 to a preset voice name (e.g. Charon, Kore) or a POST /voices user-voice id.
  • Reference-to-video: use referenceImage_1referenceImage_7 (local file paths, uploaded for you).
  • Video-to-video edit: set referenceVideo_1 to a local MP4 plus startFrameIndex_1 / endFrameIndex_1 for the trim window (output max 10 s).

Every parameter is documented on POST /videos. Local image/video paths in prompts.json (e.g. ./subject.jpeg, ./source.mp4) are inputs you supply — they are not included in this repo.


Support: Discord · Telegram · YouTube