![\"Picsart](\"https://raw.githubusercontent.com/PicsArt/gen-ai-cli/main/banner.svg\")
\n\n \n
` | ElevenLabs TTS — `en`, `fr`, `de`, … |\n| `--voice ` | ElevenLabs TTS, audio models with named voices |\n| `--rendering-speed` | Ideogram v3 / Character (`FLASH`/`TURBO`/`DEFAULT`/`QUALITY`); Kling video T2V/I2V and Kling Avatar (`std`/`pro`) |\n| `--video-id ` | Chained ops that reference an existing generation |\n| `--remove-bg-noise` | ElevenLabs Speech-to-Speech (`eleven-sts-v2`, `eleven-multilingual-sts-v2`) |\n| `--source-image-id ` | `recraft-explore-similar` (variations of an existing asset) |\n| `--similarity <1-5>` | `recraft-explore-similar` |\n\n```bash\n# Reproducible image — same prompt + seed = same result\ngen-ai generate -m flux-2-pro -p \"a red cat\" --seed 12345\n\n# ElevenLabs TTS with explicit language\ngen-ai generate -m eleven-v3 -p \"Hello world\" --language en\n\n# Recraft Explore Similar — generate variations of an existing asset\ngen-ai generate -m recraft-explore-similar --source-image-id abc123 --similarity 4\n\n# ElevenLabs STS with background-noise removal\ngen-ai generate -m eleven-sts-v2 --audio voice.mp3 --remove-bg-noise\n```\n\n---\n\n## Piping & scripting\n\nThe CLI is pipe-aware:\n\n```bash\necho \"logo for a coffee shop\" | gen-ai generate -m flux-pro -s\n```\n\nCombined with `--json`, it composes well with `jq`:\n\n```bash\ngen-ai generate -m flux-pro -p \"tabby cat\" -s --json | jq -r '.url'\n```\n\n`--no-input` prevents any interactive prompt — fail fast in CI rather than\nhanging on a missing flag.\n\n---\n\n## Environment variables\n\n| Variable | Purpose |\n|---|---|\n| `GEN_AI_API_URL` | Override the API endpoint |\n| `GEN_AI_UPLOAD_URL` | Override the upload endpoint |\n| `PICSART_CLI_VERSION` | Build-time version string (set by tsup) |\n| `NO_COLOR` | Standard NO_COLOR — disables ANSI when set |\n\nExample: point the CLI at an alternate environment for one run, without\nchanging your saved config:\n\n```bash\nGEN_AI_API_URL=https://api.example.com \\\nGEN_AI_UPLOAD_URL=https://upload.example.com \\\ngen-ai generate ...\n```\n\nBoth must be HTTPS (`http://localhost` is allowed for local development).\n\n---\n\n## Local data\n\nEverything the CLI persists lives under `~/.gen-ai/`:\n\n| File | Purpose |\n|---|---|\n| `credentials.json` | OAuth tokens for production (mode 600) |\n| `credentials..json` | OAuth tokens for a non-production `GEN_AI_API_URL` (mode 600) |\n| `config.json` | User preferences set via `config set` |\n| `history.json` | Generation history (capped at 500 entries) |\n| `recent-files.json` | Recently used input files |\n| `device-id` | Anonymous UUIDv4 for analytics |\n| `update-check.json` | Cached version-check timestamp |\n| `debug.log` | Crash reports (only written on unexpected errors) |\n\nTo fully reset: `rm -rf ~/.gen-ai/`.\n\n---\n\n## Analytics & privacy\n\nThe CLI sends usage telemetry alongside API requests. It contains:\n\n- `version` — CLI version\n- `app_session_id` — UUIDv7, regenerated each time `gen-ai` starts; one ID covers a full REPL session\n- `app_device_id` — UUIDv4, generated once and stored at `~/.gen-ai/device-id`\n- `country_code`, `timezone`, `locale_code` — derived from your system locale\n- `user_id` when the CLI is authenticated\n\nWe never send file contents, prompts, or generated results to anywhere other\nthan the Picsart API endpoint you'd already be calling. Set `PULSE_OPT_OUT=1`\nto disable analytics.\n\n---\n\n## Shell completion\n\nBash:\n\n```bash\ngen-ai completion bash > ~/.local/share/bash-completion/completions/gen-ai\n```\n\nZsh:\n\n```bash\ngen-ai completion zsh > \"${fpath[1]}/_gen-ai\"\n```\n\nFish:\n\n```bash\ngen-ai completion fish > ~/.config/fish/completions/gen-ai.fish\n```\n\n---\n\n## REPL mode\n\nRun `gen-ai` with no arguments to enter the interactive REPL — a numbered\nmenu, fuzzy model picker, image preview support (iTerm2 / Kitty), and\nclipboard-aware file input. Type `help` for help and `quit` to exit.\n\n---\n\n## Troubleshooting\n\n**\"Token refresh failed\"** → run `gen-ai login` again.\n\n**\"Insufficient credits\"** → `gen-ai credits` to check; top up at\n[picsart.com](https://picsart.com).\n\n**Hanging on input** → add `--no-input` to error out instead of prompting.\n\n**Crash with no detail** → re-run with `--debug`, then check\n`~/.gen-ai/debug.log`.\n\n**Want to see the exact request being sent** → set\n`GEN_AI_API_URL=http://localhost:PORT` and point at a local proxy.\n\n---\n\n## Architecture\n\nThe CLI uses a 5-layer pipeline (entry → commands → resolvers → execution →\noutput) with strict import boundaries enforced by `dependency-cruiser`. See\n[ARCHITECTURE.md](./ARCHITECTURE.md) for the full layout.\n\nModels live in `@picsart/ai-sdk`, published from the external `pa-gen-ai-sdk`\nrepo, not in the CLI itself — adding a model there and bumping the pin here\nmakes it appear in `gen-ai` automatically.\n\n---\n\n## Links\n\n- Picsart: https://picsart.com\n- API docs: https://docs.picsart.io\n- Issues: https://github.com/PicsArt/gen-ai-cli/issues\n\n## License\n\nMIT © Picsart\n","readmeFilename":"README.md"}