Open source CapCut alternative.
TypeScript
DATABASE_URL="postgresql://postgres:[PASSWORD]@[PROJECT-REF].pooler.supabase.com:6543/postgres?pgbouncer=true&connection_limit=1"
DIRECT_URL="postgresql://postgres:[PASSWORD]@db.[PROJECT-REF].supabase.co:5432/postgres"
BETTER_AUTH_SECRET="replace-with-a-long-random-secret"
# ChatGPT OpenID ID-token signing: PKCS#8 PEM RSA private key (2048+ bits).
# Generate outside the repository: openssl genpkey -algorithm RSA -pkeyopt rsa_keygen_bits:3072
CHATGPT_OIDC_PRIVATE_KEY="replace-with-private-key-pem"
# Vercel sends this as a bearer token on cron invocations; the cron routes
# (analytics run, cut-cloud gc) reject requests without it.
CRON_SECRET="replace-with-a-long-random-secret"
# Bearer token /api/blog/revalidate requires before it clears and re-renders
# the cached blog pages. Whatever writes blog posts sends the same value.
BLOG_REVALIDATE_SECRET=
GOOGLE_CLIENT_ID="replace-with-google-web-client-id"
GOOGLE_CLIENT_SECRET="replace-with-google-web-client-secret"
# Optional hosted inference providers.
ELEVENLABS_API_KEY="replace-with-elevenlabs-api-key"
# Cut's background-removal matte segmenter.
FAL_KEY="replace-with-fal-api-key"
# Required for Gemini / Google Cloud inference through Vertex AI. The adapter
# uses the global endpoint and derives the project from project_id.
GOOGLE_APPLICATION_CREDENTIALS_JSON='{"type":"service_account",...}'
# Dev-only: lets the eval scripts and a hand-run worker call the dev server
# with the x-donkey-dev-auth-bypass header (no session, no credits). Only read
# when NODE_ENV is not production; a hosted deployment never sets it.
DONKEY_DEV_AUTH_BYPASS=1
# Dev-only: a Gemini Developer API key. Production uses the keyless Vertex
# path above.
GEMINI_API_KEY="replace-with-gemini-api-key"
STRIPE_SECRET_KEY="replace-with-stripe-secret-key"
STRIPE_WEBHOOK_SECRET="replace-with-stripe-webhook-signing-secret"
STRIPE_PORTAL_CONFIGURATION_ID="replace-with-stripe-portal-configuration-id"
# PostHog analytics (client-side project key; public by design, inlined at build).
# Leave unset to disable tracking (e.g. local dev).
NEXT_PUBLIC_POSTHOG_KEY="replace-with-posthog-project-key"
# Cloudflare R2 (Cut cloud-mode media storage)
R2_ACCOUNT_ID="replace-with-r2-account-id"
R2_ACCESS_KEY_ID="replace-with-r2-access-key-id"
R2_SECRET_ACCESS_KEY="replace-with-r2-secret-access-key"
# Cut render worker wake-on-demand. Holding this secret is what turns it on; it
# must match the Worker's CUT_WAKE_SECRET. The address it posts to is code
# (CUT_WORKER_WAKE_URL in src/cut/lib/hosts.ts). Unset in local dev (run the
# worker by hand; it polls on its own).
CUT_RENDER_WAKE_SECRET="replace-with-wake-secret"
# Cloud-media delivery. Required wherever cloud projects are served: every byte
# the browser reads goes through the Cut Worker's R2 binding, addressed by a
# token signed with this. Without it, media reads answer 500. Must match the
# Worker's own CUT_MEDIA_SIGNING_SECRET (openssl rand -hex 32); the hostname is
# code, in src/cut/lib/hosts.ts.
CUT_MEDIA_SIGNING_SECRET="replace-with-media-signing-secret"
# Cut share streaming. Records which HLS ladder version a project publishes, in
# a Workers KV namespace the code creates on first write and then finds by
# title (src/cut/server/cloud/ladderStore.ts) — so this token is the whole
# setup. Needs Workers KV read+write on the account in R2_ACCOUNT_ID. Required
# wherever shares stream: without it a ladder render fails outright and the
# stream route logs and 5xxs, so a misconfiguration is visible rather than
# silent — viewers meanwhile fall back to the project's flattened proxy. The
# render worker holds the same value as a Worker secret, since the container
# writes the pointer when a ladder finishes.
CLOUDFLARE_KV_API_TOKEN="replace-with-kv-api-token"
# typesafe.ai
TYPESAFE_API_KEY="replace-with-typesafe-ai-key"