Hi HN, I built PoseForge because [one sentence: e.g. "getting the same character into a specific pose with AI meant endless prompt tweaking, and my photos ended up on someone else's servers"].
You give it a character photo and a pose reference, and it generates the same character (or up to four characters) in that pose. Characters are saved in a library, so you can reuse one later with a new pose photo.
It's local-first: projects and assets stay on your machine, and there's no account. It runs with the Codex CLI or Google Antigravity CLI (using a subscription you already have), ComfyUI, or hosted models through your own API keys (OpenAI, Gemini). The embedded database means no Docker or Postgres setup.
To try it:
git clone https://github.com/vishwakulkarni/PoseForge.git
cd PoseForge
npm run setup
I've tested it on [macOS / your OS]. [Windows/Linux: untested / works / unknown].
Known limits: [one or two honest ones, e.g. "complex or partly hidden poses can fail, and likeness can drift on some inputs"]. Video isn't built yet. The title says "and video" because that's where I'm taking it (a local AI film-making studio with consistent characters across shots), but today it's images only.
Please use photos you have the right to use.
I'd most like to hear where it breaks. Happy to answer anything.
Hi HN, I built PoseForge because [one sentence: e.g. "getting the same character into a specific pose with AI meant endless prompt tweaking, and my photos ended up on someone else's servers"].
You give it a character photo and a pose reference, and it generates the same character (or up to four characters) in that pose. Characters are saved in a library, so you can reuse one later with a new pose photo.
It's local-first: projects and assets stay on your machine, and there's no account. It runs with the Codex CLI or Google Antigravity CLI (using a subscription you already have), ComfyUI, or hosted models through your own API keys (OpenAI, Gemini). The embedded database means no Docker or Postgres setup.
To try it:
I've tested it on [macOS / your OS]. [Windows/Linux: untested / works / unknown].Known limits: [one or two honest ones, e.g. "complex or partly hidden poses can fail, and likeness can drift on some inputs"]. Video isn't built yet. The title says "and video" because that's where I'm taking it (a local AI film-making studio with consistent characters across shots), but today it's images only.
Please use photos you have the right to use.
I'd most like to hear where it breaks. Happy to answer anything.
[flagged]