Sora and Kling workflows emerge

New text‑to‑video workflows — examples named Sora and Kling — chain reference clips, lighting/camera prompts, and ElevenLabs audio to speed cinematic outputs. (x.com) Creators are using a 'sub/scene' tool to map a reference clip into lighting and camera prompts that other models can follow. (x.com)

A new AI video workflow is taking shape: creators are breaking a reference clip into shot-by-shot camera and lighting instructions, then rebuilding it in tools such as Sora and Kling. (developers.openai.com) (app.klingai.com) The basic idea is simple: use one clip as a visual template, extract the parts that matter, and feed those parts back into a generator as prompts. OpenAI’s Sora docs tell users to specify subject, motion, lighting, style, and camera direction, and to iterate because small prompt changes can shift the result. (help.openai.com) (developers.openai.com) Sora’s current tools support image references, reusable character assets, clip extension, editing, and batch rendering through its application programming interface. OpenAI says Sora 2 and Sora 2 Pro support 16- and 20-second generations, with Sora 2 Pro offering exports up to 1920 by 1080 or 1080 by 1920. (developers.openai.com 1) (developers.openai.com 2) Kling is exposing a similar production logic from the other side: storyboarded image-to-video generation. Its developer documentation says users can define up to six storyboard segments, each with its own prompt and duration, which makes it easier to map one sequence into several controlled shots. (app.klingai.com) Audio is becoming another modular step in the same pipeline. ElevenLabs says its text-to-speech tools generate spoken audio with controllable pacing and expression, with models ranging from low-latency Flash v2.5 at about 75 milliseconds to Eleven v3 for more expressive delivery. (elevenlabs.io 1) (elevenlabs.io 2) That means creators no longer need one model to do everything in one pass. They can use a reference clip to lock in framing and movement, a video model to render the visuals, and a separate voice model to add narration or dialogue. (developers.openai.com) (elevenlabs.io) The shift is showing up now because the tools have added more of the controls editors already use. Sora’s current app lets users trim, reorder, stitch, extend, reprompt, and remix clips, while Kling’s storyboard parameters formalize shot-by-shot planning instead of relying on one long text prompt. (help.openai.com) (app.klingai.com) The tradeoff is that these systems still require heavy iteration. OpenAI’s prompting guide says the same prompt can produce different results on different runs, and its help center says characters and voices are not always consistent today. (developers.openai.com) (help.openai.com) OpenAI has also put a date on part of this stack: its developer docs say the Sora 2 video generation models and Videos application programming interface are deprecated and will shut down on September 24, 2026. Even so, the workflow those tools helped normalize — reference in, shot plan out, voice layered on later — is spreading across AI video production. (developers.openai.com)

Get your own daily briefing

Scout delivers personalized news, insights, and conversations tailored to your role and industry.

Download on the App Store

Shared from Scout - Be the smartest in the room.