Recut

Recut is an AI agent that plans and runs multi-step generative jobs across many models. You describe what you want and attach photos, clips, audio or 3D files; it breaks the job into steps and wires each result into the next (keyframe to clip, clips plus narration and music into one cut). For each step it shows the strongest models side by side from the fal and Replicate catalogues and ElevenLabs (Kling, Veo, Seedance and others) with the cost of a run, and you approve before it runs. Covers image generation and editing, upscaling, text/image-to-video, lipsync, talking avatars, dubbing, voiceover, music, sound effects, transcription and image/text-to-3D, plus everyday edits like captions, trimming, merging, 9:16 resizing and loudness, many run on-device. Every input and output stays on a timeline to compare versions. No API keys needed (Recut credits, from $5), or bring your own provider keys. Web, Mac, Windows (beta) and Android.