Playbooks / Set up the image generation and manipulation toolkit (Gemini images, resize and crop, AI upscale, shared library)

Set up the image generation and manipulation toolkit (Gemini images, resize and crop, AI upscale, shared library)

Set up the image generation and manipulation toolkit (Gemini images, resize and crop, AI upscale, shared library)

Ready now

Generate specific visuals with the Gemini image model through curl, manipulate images (resize, crop, alpha, compress), upscale with Real-ESRGAN locally, and keep a shared indexed image library; never AI-generate a real person's face.

Kit

No separate kit. The playbook is self-contained.

You need

AccountWhyFree or paidPlan
Google AI Studio (Gemini API)Image generation (gemini-2.5-flash-image) for slides, blog images and placeholders; video analysis in the n8n research agent; the external second-eyes review of video rounds.Free tier availableFree tier works (rate limited, roughly 5 to 15 requests per minute). Add billing for volume image generation.
Python 3.12 + uvRuns every pipeline script (video gates, library export, research cache, Brain export, PDF generator, MIDI tools) inside per-project virtual environments.FreeFree.
Homebrew (macOS package manager)Installs ffmpeg, yt-dlp, gh, node, python, uv, pdftoppm (poppler) and WeasyPrint dependencies. Every local pipeline assumes it.FreeFree, open source. Apple Silicon Mac assumed (VideoToolbox encoding, mlx-whisper).
Real-ESRGAN (AI image upscaler, ncnn Vulkan binary)Adds real detail when an image must be enlarged past its native size. Lanczos and ffmpeg resizing cannot add detail; marketing graphics upscaled that way look soft.FreeFree, open source. Runs locally on Apple Silicon through Vulkan; no GPU cloud needed.
ffmpeg + ffprobeEvery cut, mux, loudness pass, proxy, contact sheet and platform encode. The delivery encoder uses h264_videotoolbox on Apple Silicon.FreeFree (Homebrew build includes VideoToolbox).

What you bring

  • Your brand colours and fonts
  • Your canonical headshot

The steps

  1. Set the key and prove one generation
  2. Build the shared library and its rule
  3. Install the manipulation CLI
  4. Install the AI upscaler and the canonical pipeline
  5. Write the prompt-engineering guide for blog and deck images

Get the full playbook

Part of: Install your social media posting suite