agentspace-so

Converts a reference image into a spritesheet and pet.json config, ready to drop into your Codex pets folder. Your custom pet appears alongside the built-ins next time you open Codex.

Turns a style prompt and structured lyrics (verse / chorus / bridge) into a full 44.1 kHz stereo track. Generates in seconds. Outputs downloadable .wav files.

Uses GPT Image Edit to retouch, remove, or refine parts of an image while preserving the rest. Includes prompt patterns that extract sharp results from the model's actual capabilities.

Uploads an image and automatically picks the best image-to-video model for your use case—whether you need smooth motion, audio sync, or identity preservation—then renders it.

Uses Nano Banana 2 to recompose photos—swap backgrounds, tweak lighting, add objects—while keeping the main subject sharp and recognizable. Handles batch edits up to 20 images.

One command-line tool that runs hundreds of AI models—from Stable Diffusion to video generation to face-swap—without managing servers or GPU quotas.

Describe what you want changed — background, tone, pacing, identity — and it picks the right model from a catalog and applies the edit while preserving your original motion and intent.

Removes unwanted objects (wires, watermarks, people) that appear across multiple video frames, or replaces a region with AI-generated content that matches the motion and lighting.

Routes your prompt to the best open or commercial video model (HappyHorse, Kling, Seedance, Wan), handles audio sync, and returns a downloadable MP4 without model selection overhead.

Takes a reference video or pose sequence and transfers the motion onto a target character, keeping pose intact while swapping appearance. Works with audio-driven animation too.

Routes requests to specialized pipelines: audio-driven character animation with identity swap for video, precise face replacement for stills using reference images. Runs on RunComfy.

Creates polished product shots, social cards, and logos with accurate embedded text and typography—no separate design tool needed. Routes to Flux for photorealism.

Generates a video from a text prompt using HappyHorse 1.0, handling multi-shot scenes, character consistency across clips, and synchronized audio in 6 languages.

Turns a text prompt or still image into a multi-shot video with synced audio and consistent characters. Covers all six Kling 3.0 modes including transitions and style control.

Adjusts light direction, warmth, and mood in a still image without changing the subject. Uses a dedicated relighting model that preserves the product and only shifts the light setup.

Uploads a source video and a description of what happens next. Returns an AI-generated clip that seamlessly continues the action, motion, or scene using Google Veo 3.

Takes a vertical phone video and extends it horizontally (or vice versa), filling in believable background so the central action stays intact and the whole frame looks composed.

Converts text or audio into a lip-synced avatar video using open models (OmniHuman, Wan 2-7, HappyHorse). Routes the job to the right model based on your input and output needs.

Routes your music request to the best available model—ElevenLabs for high-fidelity vocal tracks or open-weight alternatives for tag-driven composition—and returns a playable file.

Uses Flux Kontext to recompose, remove, or replace parts of an image without degrading the rest. Includes the model's documented prompting patterns for sharper results than generic image prompts.

Describe a vibe (lo-fi synthwave, upbeat pop with vocals, ambient) and get a 30-second to 4-minute stereo track. Control instrumentation, add multilingual lyrics with verse/chorus markers.

Calls Google's Nano Banana 2 image model with built-in prompting techniques that pull sharper, more consistent output than naive requests — works via RunComfy integration.

Generate images with Flux 2 Klein (Black Forest Labs' distilled fast variant of Flux 2) on RunComfy — bundled with the model's documented prompting patterns so the skill gets sharper output than naive prompting against the same model. Documents Flux 2 Klein's strengths (sub-seco…

Takes up to 9 images, 3 videos, and 3 audio tracks and generates a 4–15 second cinematic video with synchronized audio, natural lip-sync, and smooth motion.

Describe what you want changed in an image — remove objects, rewrite text, swap backgrounds, adjust colors — and it routes to the best model and runs the edit for you.

Select a region of an image and describe what should replace it. Uses mask-driven inpainting to intelligently fill the masked area while preserving the rest.

Takes a portrait image and an audio file, then syncs the mouth movement to speech with state-of-the-art accuracy. Routes through multiple engines for quality and speed tradeoffs.

Expands a photo or graphic outward in any direction to fill a new aspect ratio or canvas size. The AI inpaints what would naturally exist beyond the original edges.

Routes your image request to the right model (FLUX, Qwen, Seedream, or GPT-4V) based on speed, quality, and budget, then delivers the result edited and ready to use.

Converts a written scene, storyboard images, and optional audio into a smooth, lip-synced video clip using Wan 2.7. Handles duration, resolution, and aspect ratio.

Spins up image generation (text-to-image, editing, style transfer) inside Claude using your ChatGPT Plus subscription. No API keys, no per-image charges.

One command exposes a folder as a live web UI where you watch logs, code, and outputs stream as your agent runs — edit files in place, add comments, no sync needed.

Queries skills.sh, clawhub.ai, and GitHub in parallel for a skill by name or use case, ranks results by native metrics (installs, stars), and security-scans the top candidates' actual SKILL.md files.