# Configuration variables

> Every environment variable of vukem-video, with its default.

Source: https://alexgalhardo.github.io/vukem-video/docs/reference/configuration/

All variables are optional.

| Variable | Default | Meaning |
| --- | --- | --- |
| `VUKEM_VIDEO_HOME` | `~/.vukem-video` | workspace directory |
| `VUKEM_VIDEO_ADDR` | `127.0.0.1:7878` | bind address; a non-loopback address requires network mode |
| `VUKEM_VIDEO_MODE` | `local` | `local` or `network` |
| `VUKEM_VIDEO_FFMPEG`, `VUKEM_VIDEO_FFPROBE` | discovered | explicit binaries |
| `VUKEM_VIDEO_ENCODER` | automatic | force an H.264 encoder by FFmpeg name |
| `VUKEM_VIDEO_ALLOW_USER_X264` | `false` | allow `libx264` when the user's own FFmpeg has it |
| `VUKEM_VIDEO_IMPORT_ROOTS` | none | in network mode, directories from which import by path is allowed |
| `VUKEM_VIDEO_MAX_EXPORTS` | `2` | concurrent exports per workspace |
| `VUKEM_VIDEO_MAX_IMPORT_BYTES` | `8589934592` (8 GiB) | largest file accepted by import |
| `VUKEM_VIDEO_TRANSCRIBE_URL`, `VUKEM_VIDEO_TRANSCRIBE_KEY`, `VUKEM_VIDEO_TRANSCRIBE_MODEL` | none | OpenAI-compatible transcription API; when unset, the local provider is used |
| `VUKEM_VIDEO_WHISPER_MODEL` | `small` | local model name: `tiny`, `base`, `small` or `large-v3-turbo` |
| `VUKEM_VIDEO_TEST_TRANSCRIBER` | none | test only: `stub` selects a provider that answers every audio with four fixed words (`hello` 100-400 ms, `brave` 400-700, `new` 1100-1400, `world` 1400-1900, language `en`). Honoured only with `APP_ENV=test` in local mode; with any other `APP_ENV`, in network mode, or with any other value the server refuses to start. Never set in production |
| `VUKEM_VIDEO_MAX_EXPORTS_PER_KEY` | `1` | concurrent exports running per actor (two more may wait) |
| `VUKEM_VIDEO_MAX_INGESTS` | `2` | concurrent ingests per workspace |
| `VUKEM_VIDEO_MAX_TRANSCRIBES` | `1` | concurrent transcriptions per workspace |
| `VUKEM_VIDEO_MAX_TRANSCRIBES_PER_KEY` | `1` | transcriptions of one actor (key or user) that are queued or running; the same rule as `MAX_EXPORTS_PER_KEY` |
| `VUKEM_VIDEO_MAX_IMPORT_SECONDS` | `14400` | longest source accepted by import |
| `VUKEM_VIDEO_MAX_IMPORT_PIXELS` | `33177600` | largest source frame (7680 x 4320); no side above 8192 |
| `VUKEM_VIDEO_MAX_TRANSCRIBE_SECONDS` | `7200` | longest audio per transcription job |
| `VUKEM_VIDEO_MAX_EXPORT_BYTES` | `4294967296` (4 GiB) | largest exported file |
| `VUKEM_VIDEO_MIN_FREE_BYTES` | `2147483648` (2 GiB) | writes are refused below this free space on the workspace volume |
| `VUKEM_VIDEO_CACHE_MAX_BYTES` | `2147483648` (2 GiB) | size of the frame and bitmap caches before eviction |
| `VUKEM_VIDEO_RATE_LIMIT_PER_MINUTE` | `300` | requests per minute per key or local token |
| `VUKEM_VIDEO_RATE_LIMIT_SESSION_PER_MINUTE` | `1800` | requests per minute per browser session |
| `VUKEM_VIDEO_RATE_LIMIT_HEAVY_PER_MINUTE` | `10` | calls of `JobStart`, `MediaImport`, `TemplateImport`, `ProjectBundleExport`, `ProjectBundleImport`, `FontImport` and `ProjectOtioImport` per minute per actor |
| `VUKEM_VIDEO_SANDBOX` | `none` | `bwrap` runs FFmpeg and whisper.cpp in a sandbox on Linux |
| `VUKEM_VIDEO_INSECURE_ALLOW_ROOT` | `false` | allow network mode as root |
| `VUKEM_VIDEO_ADMIN_EMAIL`, `VUKEM_VIDEO_ADMIN_PASSWORD` | none | network mode: create the first administrator at start-up when no user exists |
| `VUKEM_VIDEO_URL`, `VUKEM_VIDEO_KEY` | none | client side: make the CLI and the stdio MCP server talk to a remote server with an API key |
| `VUKEM_VIDEO_PROJECT` | none | client side: the default project of CLI commands that take one |
| `OTEL_EXPORTER_OTLP_ENDPOINT` | none | turns on OpenTelemetry (ADR-0026) |
