Step1X-Edit v1p2
Free.ai (self-hosted)
·
image_edit
·
~1000 tokens per edit
Step1X-Edit v1p2 is an image editing model built by StepFun. Strongest at Reasoning image edits: multi-step instructions, referential prompts, hand / anatomy fixes, abstract style transfers. Pair with Qwen-Image-Edit (faster, direct) for the best of both worlds. Apache 2.0.. Self-hosted on Free.ai GPUs — runs free against your daily token pool (1,000 tokens per edit). Released under Apache 2.0 — commercial use permitted on Free.ai.
Use via API
OpenAI-compatible REST API. Generate a key and call this model in seconds.
curl -X POST https://api.free.ai/v1/image/edit/ \
-H "Authorization: Bearer sk-free-..." \
-H "Content-Type: application/json" \
-d '{"model":"step1x-edit","prompt":"your prompt here"}'Similar models
Frequently Asked Questions
Step1X-Edit v1p2 is one of the AI image-editing models on Free.ai. Upload a source image, describe the edit you want in the prompt, and Step1X-Edit v1p2 returns a new image. Inpainting, style swaps, object removal, and text-driven edits are all in scope.
Strongest on whole-image style changes, object swaps with clear bounds, and lighting/color shifts. Weaker on tiny fixes inside cluttered scenes — for surgical pixel-level work try the masked-edit flow on /image/edit/.
PNG, JPG, WebP, and HEIC up to 10 MB. Smaller files render faster. Transparency is preserved on PNG output.
Step1X-Edit v1p2 preserves overall composition by default — the same subject in the same pose, with your edit applied. To deviate further, use a more directive prompt or lower the "preserve" strength in advanced settings.
Step1X-Edit v1p2 runs on our own GPUs and draws from your daily free pool first; once depleted, paid token packs start at $5 → 200,000 tokens. About ~500 tokens per edit, minimum 100.
Free.ai will not strip a copyright notice or third-party watermark you do not own — both the upload check and the model refuse. You can remove your own watermark or any element you own the rights to.
Same resolution as the source up to 2048×2048 max. Larger sources are auto-downscaled; we keep the aspect ratio so the result frames cleanly back over the original.
Yes — /batch/ accepts a folder of images + one edit prompt. Each output saves to /account/?tab=history. For per-image prompts, queue them through the API.
Yes — POST your image + prompt to /v1/image/edit/ with model="Step1X-Edit v1p2" (or the slug). See /api/ for auth, error codes, and code snippets.
Self-hosted models stay on Free.ai GPUs. Premium models pass your image to the upstream provider under our DPA. We do not train on your inputs and do not sell data. Uploads are deleted after the share-window (24h anon, 7d signed-in).
Yes — Free.ai grants commercial use of the output. You must own (or have license to) the source image you upload; we cannot grant rights we do not have.
Yes — failed jobs auto-refund to the source (daily pool or paid tokens). If a refund did not land same-day, email contact@free.ai.