Sora 2 Pro storyboard after the API

The 10 / 15 / 25 second Sora 2 Pro storyboard row dies with the API on 24 September 2026. It is image-to-video, 30 credits, native audio. Replacement is not “find another storyboard.” Replacement is: pick by duration and whether you still have the stills — Seedance 2.0 / 2.5, Flux 3 T2V or first-last, Veo first-last, or movie-mode on Versely.

Sora 2 Pro Storyboard is the catalog card: category image-to-video, durations 10s / 15s / 25s, credits 30, audio true, resolution unset. Do not invent 4K for it. Do not treat it as the 4 / 8 / 12 / 16 / 20s Sora 2 Text to Video row. Different input, different ticks, same OpenAI shutdown.

OpenAI’s Help Centre dates the API discontinuation to 24 September 2026 (sora-2, sora-2-pro, snapshots). Export stills and keepers at sora.chatgpt.com/sunset before deletion. This page is only what to run instead of Pro storyboard.

What the dying row actually was

Pro storyboard was the I2V door with a board: you brought frames, you asked for 10, 15, or 25 seconds, and native audio came back with the picture. The still was the contract. The length was longer than Veo’s 8-second ceiling. Twenty-five seconds is longer than Seedance 2.0’s 15-second ceiling.

After 24 September 2026 you cannot plan production on that combination. You can keep files you already have. You cannot assume a new 25-second I2V generate from the same slug.

Two mistakes show up on the 25th:

• Prompting T2V as if it were the board. Text-to-video invents the frame. If legal signed a pack still, T2V is a cousin-maker unless you are only drafting.
• Asking Veo 3.1 for 25 seconds. Veo 3.1 is 4 / 6 / 8s, 40 credits, 4K, native audio, text-to-video. First-last is 4 / 6 / 8s. Neither is a 25-second storyboard.

The stills you used on the board are the most valuable leftover. Export them. Name them. They decide which replacement row is legal.

Pick by duration and whether you have stills

Write two facts on the job: seconds and inputs.

You have stills, and 4–8 seconds is enough. Veo first-last is 40 credits, image-to-video, 4 / 6 / 8s, 4K, native audio, requires the frames. Use it when the opening and closing plates are already true and the job is a short interpolation with sound. It will not give you 10s, 15s, or 25s.

You have stills, and 5–20 seconds is the band. Flux 3 first-last is 9 credits, 5–20s, native audio, requires frames. That is the length overlap with 10s and 15s Pro jobs, and it can take a 20s ask. It is two frames, not a five-still board. If you had a stack of references rather than a first and a last, you are on a reference-to-video row instead.

You have a reference stack, and 4–15 seconds at 1080p is enough. Seedance 2.0 is 69 credits, text-to-video and reference-to-video, 4–15s, 1080p, native audio. This is the SKU-hold landing when identity has to survive a turn. Seedance 2.0 Fast Reference to Video is 25 credits, 720p, 4–15s, native audio — the cheaper ref draft. Fast is not 1080p. Fast is not 25 seconds.

You have a prompt (and maybe stills via the image-to-video tool), and the shot must pass 15 seconds. Seedance 2.5 is 47 credits, text-to-video, 4–30s, 720p, native audio. ByteDance announced the line at Volcano Engine FORCE on 23 June 2026. On Versely the ceiling is 720p, not a 4K claim from another host. A 25-second Pro file maps here on length. It does not map here on “the still is guaranteed” unless you drive it from a still through AI image-to-video and inspect identity.

You have no stills, and you want 5–20 seconds with native audio. Flux 3 Text to Video is 9 credits, 5–20s, native audio. That is a generate, not a board. Use it when the Pro job had become prompt-led anyway.

You need a chained 30–60 second piece with scene handoffs. That is not a single model tick. Movie Mode is a deliverable shape: planned scenes, last-frame handoffs, one export. The tool door is AI movie maker. Use it when Pro storyboard was standing in for a multi-scene film. Do not use it for a 10-second pack turn.

If the Pro job was 25 seconds and you refuse 720p, you do not have a 1080p 25-second Seedance 2.0 card. Seedance 2.0 stops at 15s. Flux 3 first-last and Flux 3 T2V stop at 20s. You either cut, accept 720p on Seedance 2.5, or chain.

What not to use as a fake storyboard

Kling 3 Turbo is 12 credits, 3–15s, 1080p, audio false. It will not carry a 25-second board. It will not carry native audio. It is a silent draft row.

Wan 2.7 I2V is 15 credits, 2–15s, 1080p, and audio is null on that catalog row. It can animate a still. It is not a native-audio storyboard replacement. If you need sound born with the picture, stay on Seedance, Flux 3, or Veo.

Gemini Omni Video is 9 credits, T2V / I2V / R2V, 4K, 4 / 6 / 8 / 10s. Audio is null on the Versely card. It can take a 10-second generate. It cannot take 15s or 25s. Do not invent a /models/gemini-omni-1-1-flash slug for Google’s 27 August 2026 Omni 1.1 Flash surface.

Do not prompt “storyboard” as a magic word on a T2V row and expect Pro’s still-lock. The word does not attach the JPEG.

Export first if the stills only live in Sora. The export page is the file job. Without stills, every replacement that needed I2V becomes T2V with drift.

A working replacement sequence

1. Export stills and the 10 / 15 / 25s keepers.
2. Label each job with duration and input: one still, first+last, ref stack, or prompt-only.
3. Match the row with the matrix above. One row per job.
4. Inspect identity before you caption. If the pack lettering slipped, do not extend the miss. Change row or change input.
5. Caption and disclose on the file you will upload. Add captions to video. Article 50 does not pause because the board moved hosts.

If you have a 15-second Pro keeper you still like, you may not need a replacement generate at all. Ship the file. Use a new row only for the next shot. Regenerating approved footage “so it isn’t Sora” is how identity dies twice.

When the stills are on disk, run the replacement row from Versely.

FAQ

What dies on 24 September 2026 for this row?

The OpenAI Sora API, including sora-2-pro. The Versely card Sora 2 Pro Storyboard is 10 / 15 / 25s, I2V, 30 credits, native audio. After the API is gone, do not plan new jobs on that slug. Keep exported files.

Which row replaces a 25-second storyboard?

On length, Seedance 2.5 (4–30s, 720p, 47 credits, native audio) is the single-pass ceiling that clears 25s. Flux 3 stops at 20s. Seedance 2.0 stops at 15s. Veo first-last stops at 8s. If 720p is illegal, cut the job or chain scenes in movie-mode instead of hunting a 1080p 25s twin.

Can I replace Pro storyboard with text-to-video only?

Only if the still was never the contract. If a pack, face, or set still was approved, use I2V, first-last, or R2V. Flux 3 T2V and Seedance 2.5 T2V will invent a frame. That is a different job.