Describe the job — dub a video, swap a face, clean a podcast, cut shorts. Shape it as nodes on a canvas. Let it run on real GPUs. Review the cut in your gallery.
Start free — $1 creditFree creations are watermarked and join the public showcase. Paid plans keep everything private.
Say what you need in plain language, or start from a ready-made template.
Every job is a graph of nodes on a canvas — rewire, swap models, tune parameters.
Steps queue to GPU workers running open-source models. Watch each node light up live.
Outputs land in your gallery with full provenance — re-run, tweak, share.
Translate speech, re-voiced in the original speaker's voice, lips synced.
FaceFusion across a whole video, face restoration included. Consent-gated.
Find the best moments in long footage, cut vertical clips with captions.
Prompt to clip with modern open-source video models.
Denoise, strip filler words, normalize loudness. Background replacement for audio.
Describe a track; get a score. Stems and vocals included.
Cut out, relight, and place products into generated scenes.
Describe what to remove; segmentation plus inpainting does the rest.
OCR any PDF or scan into validated, structured JSON.
Speaker-labeled transcripts with minutes and action items.
One prompt → script → scenes → narration → score → assembled film.
33 operations. Draw any pipeline the catalog can express.
Everything the canvas does, a key can do: run workflows, poll executions, upload media, manage pipelines programmatically. Webhooks call you back when the job is done.
curl -X POST https://inferencebrain.com/v1/workflows/5/run \ -H "X-API-Key: ib2_…" \ -d '{"inputs": {"audio": "…"}}' {"execution_id": "…", "status": "pending", "estimate": 0.021}