{{site_title_here}}

Tag: tech-foolishness

  • Gemini (more than) edits your photos

    Gemini (more than) edits your photos

    Last Saturday night I had a blast photographing the Roseville Big Band perform at the Commemorative Air Force Fall Bombers Moon Ball. Trying to get more media for the Minnesota Mandolin Orchestra, I thought I’d ask Glen, who performs with both ensembles, if we could swap some photo ops.

    Context set, I was nervous during the shoot. I had walked the hanger during the day, but I wasn’t sure about lighting for the night of the shoot. Also, I had misunderstood the timing, expecting to have some time with the band ahead of the performance for a portrait. Actually, that was accurate, but I hadn’t planned for throngs of people arriving early at the hanger.

    By the time the band was ready for their photo, there were oodles of people crawling around Miss Mitchell, the aircraft that the band wanted to be photographed with. The sun was low — about 30 minutes from set and directly behind Miss Mitchell. There was a professional photographer already working with plane into his portraits whom I asked for a graceful pause.

    Unaltered photo
    Manually edited photo
    Gemini edited photo.

    Having to work quickly, I had the plan to use a 14-24mm lens and a camera-mounted fill flash and shoot close to crop out visitors. With the sky being so bright, the sun at the back of the plane, and the band dressed in black, I pushed the exposure a full step and clicked as fast as my flash would cycle.

    Thirty seconds later, I had my washed-out result, above left. To the punchline, this morning I read that Gemini can edit photos, though not NEF or TIFF today, and after several iterations, mostly of directions to brighten Gemini’s idea of a sunset and not to to distort the subject’s proportions, Google delivered the result on the right.

    If you’re not familiar with the band, it might not matter, but I doubt that many family members would recognize them in the AI-edited photo. Sandy walked up behind me and asked, “tuxedos! Is the conductor wearing the red one?” And I realized I couldn’t identify the conductor – actually, I thought he wasn’t present and I swore. Then looking at the faces, things didn’t look right. It was a hassle to brighten Gemini’s result, but I could not persuade Gemini to spare the band face lifts.

    There was value to complaining to Gemini. The AI-altered image focused my time spent on the exposure of the original image in NX Studio by giving me a target in my imagination. The center(and banner) image is the result.

  • My Introduction to Stable Diffusion

    For a few years now, I’ve been fascinated by programmatic art. For ten years or so, there’s been a stream of projects that do something along the lines of applying an artist-inspired filter (usually Van-Gogh) to existing photos to stylize them. Last year, the DALL-E project went a step further by generating images from textual descriptions and a library of images.

    I was stunned at both variations on existing images and images synthesized from descriptions. This summer, generative art got a little more accessible to the masses when Stable Diffusion was released, using similar technology, but publicly available models (at least for now).

    I think the business model leverages that you can make interesting images from the public models, but that most businesses will want to tune the image text-to-image models to suit their intended message. The models require an enormous amount of energy to train — the default one has an 11 metric-ton carbon footprint — and the model propagates cultural biases present in the underlying data. Both aspects seem significant enough to motivate an entrepreneur to hire out specialists.

    Stable Diffusion is a little tricky to install on a M1 Mac — you have to checkout a particular branch of stable-diffusion, currently `apple-silicon-mps-support`, and hope that nightly PyTorch builds are stable. You also have to register with HuggingFace to download an image model, which seem to be updated frequently at this point.

    There’s currently no GPU support for Apple silicon for Stable Diffusion, so you need to be patient, but here are some samples of the Stable Diffusion imagination…. that might sound a bit like “stable genius” and artistically sometimes the images resemble another product shoveled from a stable…. but again, samples of

  • More Javascript Futzing

    Poking at my visual programming course, here’s the result of the third assignment. Consider this under a refrigerator magnet.

    The assignment objectives seem to be to introduce use of classes and animation. The new wrinkle for me was Window.requestAnimationFrame(), which we use to schedule repainting the scene to animate.