This is a Next.js project bootstrapped with create-next-app.

Getting Started

First, run the development server:

npm run dev
# or
yarn dev
# or
pnpm dev
# or
bun dev

Open http://localhost:3000 with your browser to see the result.

Admin video studios

The admin dashboard contains five separate studios: Free Video, Free Mix, Sora 2 Slow, Combined, and Video · Veo. Free Mix alternates free Pexels and Pixabay footage and keeps a local recent-use ledger to reduce repeated clips. Sora 2 Slow submits every Sora render through the OpenAI Batch API. Combined generates one 8-second Sora scene through Batch and downloads three free Pexels scenes. Video · Veo generates four 8-second portrait clips with veo-3.1-lite-generate-001 on Vertex AI, with native audio disabled so the clips use the same narration, subtitle, editing, storage, and Telegram pipeline. Paid jobs are backed by adminAiVideoJobs, a transactionally reserved daily budget, and a resumable local worker. Generated content is English-only. Provider credentials and cost controls remain server-side.

Copy the AI video values from .env.example into .env.local. Keep AI_VIDEO_PAID_GENERATION_ENABLED=false until paid generation is intentionally enabled. Combined also needs PEXELS_API_KEY; when it is omitted, the worker safely reuses the existing key from MONEYPRINTERTURBO_ROOT/config.toml. Free Mix needs both Pexels and Pixabay keys and can reuse either key from the same MoneyPrinterTurbo config when the matching environment value is omitted. Video · Veo uses the Firebase service account by default and its project ID unless VEO_PROJECT_ID is set; that account must be allowed to call Vertex AI, and the Vertex AI API must be enabled. Veo scenes omit storageUri and receive bytesBase64Encoded from Vertex so generation does not depend on Vertex service-agent access to the Firebase bucket (which otherwise fails with “Service agents are being provisioned” on fresh projects). The worker also needs ffmpeg and ffprobe on PATH (or explicit binary paths).

Start the web app and worker in separate terminals:

cd /Users/dimab/Documents/oneiro-web
npm run dev

cd /Users/dimab/Documents/oneiro-web
npm run ai-video-worker

Safe validation commands never call the paid Videos endpoint:

npm run ai-video-check
npm run ai-video-synthetic-test

At the default Sora Batch price of $0.05/second, Preview reserves $0.20, Standard reserves $1.60, and Combined reserves $0.40. At the default Veo 3.1 Lite video-only 720p price of $0.03/second, Video · Veo reserves $0.96. The authenticated enqueue API derives pricing and durations from server configuration, requires explicit costConfirmed=true, rejects requests while paid generation is disabled, and reserves both dollars and the daily job count in one Firestore transaction. The worker saves OpenAI Batch IDs and Vertex AI operation names before polling so a restart resumes the same paid submission.

You can start editing the page by modifying app/page.tsx. The page auto-updates as you edit the file.

This project uses next/font to automatically optimize and load Geist, a new font family for Vercel.

Learn More

To learn more about Next.js, take a look at the following resources:

• Next.js Documentation - learn about Next.js features and API.
• Learn Next.js - an interactive Next.js tutorial.

You can check out the Next.js GitHub repository - your feedback and contributions are welcome!

Deploy on Vercel

The easiest way to deploy your Next.js app is to use the Vercel Platform from the creators of Next.js.

Check out our Next.js deployment documentation for more details.

Admin image studios

The admin dashboard also has Image · Sora, Image · Veo, and All Images. These reuse the same OpenAI key and Vertex service account as the video studios, but they call cheaper still-image models (gpt-image-1-mini and gemini-3.1-flash-image) instead of Sora/Veo video. After each job finishes, the card and Telegram caption show the approximate cost from provider usage. Start the image worker in a separate terminal:

npm run ai-image-worker

Scheduled social publishing

The All button on a video card opens a sheet with two ways out: publish to every connected network right away, or pick one moment for all of them. YouTube takes the scheduled moment natively — the file is uploaded private and YouTube flips it public at status.publishAt. TikTok, Instagram, Facebook, Threads and Pinterest have no native scheduling, so the batch is stored on the video document for the dashboard and canonically in RTDB under social_scheduled_assets/<id>/socialSchedule. The RTDB queue is indexed by socialSchedule/scheduledAt and published by /api/cron/social-publish.

That endpoint is driven by the Firebase scheduled function socialPublishTick (every 5 minutes, region europe-west1, time zone Asia/Jerusalem), which calls it with SOCIAL_PUBLISH_CRON_SECRET as a bearer token. Store the same value in Secret Manager and the Vercel production environment. SOCIAL_PUBLISH_ENDPOINT holds the protected production URL in Secret Manager.

If a scheduled batch fails at its moment, the worker sends a Telegram message naming the video, what did go out and what did not — TELEGRAM_BOT_TOKEN and TELEGRAM_PERSONAL_CHAT_ID have to be set on Vercel for that, and without them the failure is still visible on the video card. A successful batch stays silent.

Positioner is told about a schedule the moment it is made: every scheduled platform gets a row with the status Запланировано and the planned date. The real publish later reuses the same key and updates the row to Опубликовано with the actual time, and cancelling a schedule marks it Отменено.