bild 0.0.7

Added

• New models: Black Forest Labs FLUX Video Edit (bfl/flux-tools/video-edit-v1) and FLUX Video Upscale (bfl/flux-tools/video-upscale-v1), and Sourceful Riverflow 2 Fast (sourceful/riverflow-2-fast) and Riverflow 2 Pro (sourceful/riverflow-2-pro).
• FLUX 3 Image (bfl/flux-3-image) accepts up to ten reference images.
• Colored output in the Windows console.

Fixed

• On Windows, a leading ~/ or ~\ in --output-path, --save-results, or the output-dir setting expands to the home directory.

bild 0.0.6

Added

• New models: Black Forest Labs FLUX 3 Image (bfl/flux-3-image), Recraft V4.1 Flash (recraft/recraftv4_1_flash), and xAI Grok Imagine Video 1.5 Lite (xai/grok-imagine-video-1.5-lite).
• New OpenRouter models: FLUX 3 Image, Seedream 5.0 Flash, Ming Image 0.1 Design, Ming Image 0.1 Design Layer, Recraft V4.1 Flash, and HeyGen Video 1.
• Install with Homebrew on macOS or Linux: brew install shdeen/tap/bild.

bild 0.0.5

First public release. A bild command for generating and editing images and videos from the terminal and for effective agentic use.

Features

• Generate and edit images and videos from the terminal with one command, bild, which makes the request, polls for the result, and saves the generated media to disk.
• Provider-agnostic support for descriptor-class and adapter-class providers; current support for OpenAI, Google, xAI, Black Forest Labs, Recraft, Sourceful, Kling, and OpenRouter.
• Built-in model catalog with three non-API commands, bild list, bild search, and bild info, to list and search for providers and models and provide detailed information about each. No API calls are made for catalog commands, and no API key is needed.
• Shared generation options, i.e. the same flags, such as --aspect-ratio, --resolution, --size, and --duration, work across providers. (When a model does not accept a value, Bildomat uses the nearest value that the model accepts, or omits it, and reports the change.)
• Take input media with -i to edit an image or guide a result with local files or HTTP(S) URLs in PNG, JPEG, WebP, or MP4 format.
• Allow specifying video keyframes for opening and closing frames with a first: or last: prefix to input media and, for models that support timed keyframes, with exact number of seconds.
• Output flag -o/--output-path allows one argument to set directory, filename stem, and desired format. Format is honored if the model supports it; otherwise, save file with the format returned, and report the filename.
• Ensure safe saving of media; for filenames in use, append a suffix beginning with a 2-digit '02', and report the path of every file saved.
• Output options for better scriptability and agentic use. Specifying -j/--json returns a JSON structured result, --print-filename prints only saved paths, and --save-results writes the result to a file. Exit codes distinguish usage errors from other failures.
• Save thought text from supporting Gemini image models in a Markdown file beside the image by adding --include-thoughts flag.
• Install scripts for macOS and Linux (install.sh) and Windows (install.ps1).
• Configuration file in ~/.bildomat/config.yml sets a default model, a default output directory, and provider API keys. Keys can also be set in environment variables. YAML keys and env vars are listed on each provider's info page.