Video Transcription & Multiplayer Game Services
AI-powered subtitle generation using OpenAI Whisper & NVIDIA Riva
Real-time multiplayer deception game with AI topics
Secure user auth with email/password and Google OAuth2
Browse and download AI model files via authenticated API
Convert videos to timestamped subtitles with AI transcription
Upload a video file and start transcription
Check status and progress of a transcription job
Download the generated SRT subtitle file
Retrieve a list of all transcription jobs
Delete a transcription job and its files
tiny — Fastest, lowest accuracy (~40 MB)
base — Balanced speed/accuracy (~140 MB)
small — Good accuracy (~500 MB)
medium — High accuracy (~1.5 GB)
large-v3 — Highest accuracy (~3 GB)
57+ languages including English, Spanish, French, German, Chinese, Japanese, Arabic, Hindi, and more.
file — Video file to transcribe (required)
language — Language code (e.g. en, es) — auto-detect
if omitted
translate — on or off — translate to English
model — tiny, base, small,
medium, large-v3 (default: medium)
Upload a video/audio file and queue transcription via NVIDIA Riva. Supports large files with automatic chunking.
file — Video or audio file to transcribe (required)
language — Source language code (default: ja)
translate — on or off — translate to English (default:
on)
Status, download, list, and delete reuse
the standard /api/jobs endpoints above.
Real-time multiplayer game where players guess who has a different topic
Create a new game session with a category
Join an existing game with session code
Start the game and assign imposter (minimum 3 players)
Get current game state — shows different topic based on role
Vote for suspected imposter during voting phase
End discussion and start voting phase
End voting and determine game results
Retrieve reveal/result after voting
Get list of all games waiting for players
Start a fresh round for the same players
Keep a player marked active in a session
Remove inactive players from waiting games
Delete old/waiting game sessions
Delete a game session and all its data
Waiting — Players can join the game
Discussion — 10 minutes — players discuss the topic
Voting — 1 minute — players vote for suspected imposter
Result — Reveal and scores shown
Secure user authentication with email/password and Google OAuth2
Authentication uses JWT tokens stored in HttpOnly cookies. After login, a secure cookie is set automatically—no need to manage tokens manually.
Two flows are supported: email/password registration & login, and Google OAuth2 single sign-on.
Create a new account with email and password. Uses OAuth2 form encoding.
Authenticate with email/password. Returns a JWT in a secure HttpOnly cookie.
Redirects to the Google OAuth2 consent screen
Handles the OAuth2 callback, creates/updates user, sets JWT cookie, and redirects to the frontend
Return the profile of the currently authenticated user
Clear the HttpOnly authentication cookie and end the session
Register & Login use OAuth2 form data
(application/x-www-form-urlencoded):
username — User's email address
password — User's password
Cookie name: access_token
HttpOnly: Yes — not accessible via JavaScript
Secure: Yes — HTTPS only
SameSite: None — cross-origin supported
Browse and download AI model files from the server
All model endpoints require a valid X-Api-Key header. Contact the administrator to
obtain an API key.
Return metadata for every available model file
Stream a model file as a binary download
Each model object includes:
id — Stable, URL-safe short ID (SHA-256 derived)
name — Human-readable model name
fileName — Original file name on disk
downloadUrl — Direct download path
sizeBytes — File size in bytes
lastModified — ISO 8601 timestamp (UTC)
checksum — SHA-256 hex digest for verification
curl -H "X-Api-Key: YOUR_KEY" https://api.ayush.ltd/api/models
List models: 2 requests / minute
Download model: 5 requests / hour