UpRes CLI is the official command-line tool and software development kit (SDK) for upres.ai, designed for image, video, and audio restoration scenarios. It provides 14 public aliases through a unified API, with output resolution support up to 8K.
In terms of use, UpRes CLI is mainly used to submit image, video, and audio files to upres.ai for upscaling/restoration processing and to retrieve the processing results. The official description lists its differences from Topaz, Upscayl, and other SaaS upscaling tools: Topaz is relatively expensive and desktop-only; Upscayl is suitable for single processing but difficult to support batch export or CI pipelines; other SaaS upscalers may have issues such as watermarked output, resolution limited to 2K, or per-image charging with no monthly plan. UpRes provides a clean REST API, 14 public aliases, batch processing, and predictable pricing (the official statement says Creator starts at $9 per month).
Main capabilities include: image, video, and audio restoration; output up to 8K; 14 public aliases, each of which clearly states whether it "generates details" or "does not perform additional processing on the file"; aliases that do not upscale ignore the --scale parameter; batch processing support; files over 50 MB are automatically uploaded via the TUS resumable upload protocol; a Python SDK is provided, uses httpx, supports async, and has no heavyweight dependencies. The complete model catalog can be viewed at upres.ai/models, and the live specification is at api.upres.ai/v1/openapi.json. The Studio tier unlocks the full API, batch processing, and removal of output watermarks.
In terms of usage, first register at upres.ai (free, 5 upscales per month), then generate an API key at upres.ai/account/api-keys (the key is shown only once). The key resolution order is: --api-key command-line parameter → UPRES_API_KEY environment variable → ~/.config/upres/config.json configuration file. Installation supports global installation; the official text does not provide the specific installation command content. MCP support is also provided, suitable for Claude, Cursor, and Hermes, and can be configured in claude_desktop_config.json or Cursor. MCP tools include: upres_list_models, upres_get_credits, upres_upscale_image, upres_upscale_video, upres_enhance_audio, upres_get_job.
Applicable scenarios include: developers who need to automate or batch process image, video, and audio restoration; teams that need to integrate upscaling capabilities into CI pipelines; SaaS or application developers who need API integration; and Claude, Cursor, and Hermes users who use MCP.
In terms of limitations and supplements: the official free tier is provided, with 5 upscales per month; for eligible open-source projects, you can send the repository link to support@upres.ai to apply for 500 operations per month. The SDK automatically handles some errors: in Python it throws AuthError, QuotaExceededError, or UpresError. Contribution guidelines are in CONTRIBUTING.md; the official text does not provide the specific command for running tests. The project is licensed under MIT, see LICENSE for details.
