Sora 2 to Seedance 2.0 migration

Move Sora 2 work to Seedance 2.0 for up to 12 references, phoneme lipsync in 8+ languages, and 15s clips. Credits only. The Sora app and web already died on 26 April 2026. The Sora API — sora-2, sora-2-pro, all snapshots — shuts on 24 September 2026. This page is not a generic sunset map. It is the Seedance 2.0 row: what it takes from a Sora job, what it cannot take, and how Versely meters it in credits.

What you are actually moving

Sora 2 text-to-video in the catalog is native audio, 720p, 10 credits, durations 4s, 8s, 12s, 16s, 20s. Sora 2 Pro Storyboard is native audio, image-to-video, 30 credits, durations 10s, 15s, 25s. After 24 September you cannot re-run those ids on OpenAI's API. Files you already downloaded remain files. Jobs you have not exported are the ones this week is for.

Seedance 2.0 is ByteDance's row: native audio, 1080p, 69 credits, text-to-video and reference-to-video, every second from 4s through 15s. Vendor materials (fal, Atlas Cloud, Morphic, as of 4 September 2026) add the capabilities that make it the Sora replacement for talking, referenced, mid-length work: unified multimodal input (text, image, video, audio; up to 12 references), phoneme-level lip-sync in 8+ languages, up to 15 seconds per generate. Quote those as vendor capabilities. On Versely, quote 69 credits for this slug, not a dollar figure.

The duration cut is the first honest conversation. Sora 2 T2V went to 20s. Storyboard went to 25s. Seedance 2.0 stops at 15s. A 16s or 20s Sora prompt does not port as one call. Split it, or accept 15s, or pick a different family for the long hold. Do not prompt "20s" at Seedance and hope.

Resolution goes the other way on this slug: Sora 2 T2V is listed at 720p; Seedance 2.0 is listed at 1080p. That is not a reason to upscale old Sora files. It is a reason to regenerate on Seedance when you need the new master.

Storyboard as a container does not port. Seedance will not reconstruct Sora's identity token across a twelve-beat board. Do not rebuild Sora Storyboard in one model. For Seedance, the replacement is references plus one 15s pass, or stills then motion, not a fake board button.

Multi-reference and the 15-second cap

Sora jobs that were "the same person, the same bottle, a new kitchen" should not go back to a paragraph. Seedance 2.0's point is the stack. Vendor cap is up to 12 references across text, image, video and audio. On Versely, the related fast door is Seedance 2.0 Fast reference-to-video: up to nine images, three videos, three audio inputs, 4–15s, 720p, 25 credits, native audio. Use the 69-credit Seedance 2.0 slug when you want that family's 1080p T2V/R2V row. Use the 25-credit Fast R2V slug when the job is a cheaper referenced pass at 720p.

How to port a Sora prompt that named the product in prose:

Export the stills you actually trust — pack, face, room — not a frame of a Sora take you no longer like.
Attach them. Label in the prompt which file is the product and which is the person.
Write action and camera only. Stop re-describing the label.
Set duration to 15s or less. If the Sora piece was 20s, cut the beat that was air.

A Sora storyboard that was six scenes is not six Seedance references. References are identity and look. Scenes are time. Time still has to fit in 15 seconds or in an edit of two generates. Two 8-second Seedance calls with the same stack, cut together, is a legitimate port. One 25-second wish is not.

Six reference-to-video rows that want a stack is the wider menu. This migration page stays on Seedance because of lipsync languages and the 15s ceiling, not because it is the only R2V door.

Phoneme lipsync in 8+ languages

This is the Seedance-specific reason to land here instead of on a silent motion row.

ByteDance's Seedance 2.0 materials describe phoneme-level lip-sync in eight or more languages, with native audio-visual sync in one pass. Sora 2 also returned native audio. What you are buying on the move is not "sound exists." It is mouths that track a line in more than English, on a 15-second cap, with references so the face is the face you attached.

Port rules for talking jobs:

Put the spoken line in the prompt, in the language you want, in quotes. Do not ask the model to "say something friendly."
Attach a face still if the person has to match a campaign. Phoneme lipsync on a random face is still a random face talking well.
Stay inside 15s. A Sora 20s monologue is two beats or a shorter rewrite.
Do not pick a silent row "to save credits" and then wonder where the mouths went. Seedance 2.0's published audio field is on. That is the point of this migration.

Languages: the vendor line is 8+. Do not publish a Versely language matrix that is not in the catalog. Test the line you actually need. If the take is English-only cinematic b-roll with no mouth, Seedance is optional; Veo or Kling may be the better port. If the take is a creator talking in Spanish or Hindi on a product, Seedance is the row this page is for.

Captions remain a second pass. Native audio is not burned-in type. Add captions to video after you like the take. Silent autoplay still needs pixels for the hook.

Credits, not a Sora clone

Versely prices this move in credits. Seedance 2.0 is 69 credits per generate on the published row. Fast R2V is 25 credits. Sora 2 T2V was 10 credits; Storyboard 30. The new row is not a price match. It is a capability match for referenced talking video at 15s and 1080p.

Vendor per-second dollar figures exist on fal and Atlas Cloud. Do not use them as Versely prices. They are not your bill.

What not to expect:

A 20-second single call.
Sora Storyboard's per-scene regen as a button.
A 4K guarantee on this Versely slug (the published resolution is 1080p; some hosts advertise 4K for Seedance — that is not this record).
Astra, GPT-6, or any chat model to "become" Seedance. The brain can write the prompt. The row spends the 69 credits.

Archive Sora outputs before you migrate, so you still have the old masters. Then regenerate only the jobs that have to ship after 24 September. A library of Sora files that still work as files does not need a paid remaster.

When the stack and the line are ready, run them on Seedance 2.0 from the studio. Name the slug in the log. Keep Veo and Kling for the jobs this row is the wrong length or the wrong kind of audio for.

FAQ

Can I keep a 20-second Sora prompt and just switch the model name?

No. Seedance 2.0's published durations stop at 15s. Cut the prompt or split the job. A 20s request is not a hidden setting.

Is Fast R2V the same as Seedance 2.0?

No. Fast R2V is 25 credits, 720p, reference-first. Seedance 2.0 is 69 credits, 1080p, T2V and R2V. Same family, different published rows.

Do I pay Versely in the vendor's per-second dollars?

No. On Versely this migration is credits: 69 on Seedance 2.0, 25 on Fast R2V. Third-party USD belongs on those vendors' pages, not on this bill.