Sora 2 vs Veo 3.1 vs Seedance 2.0 After Retirement

Sora 2 is no longer a current generation choice: the consumer app/web closed April 26, 2026, and the API shutdown date was September 24. For new work, compare an available Seedance route with the current Google Cloud Veo route. Google’s separate Gemini Developer API Veo 3.1 preview also shut down, on October 22.

This is a documentation-based comparison of model capabilities and ChatCut integration limits, checked October 28, 2026. It is not a matched quality benchmark, and a provider’s documented feature is not proof of access through a particular account or editor.

Last verified: 2026-10-28. No new provider generation or deployed-production test was performed for this update.

What changed since the original May comparison?

Sora’s retirement dates passed. The earlier advice to lock hero shots before September is no longer actionable. Existing exported footage can remain part of a project, but new-generation dependencies need a different available route.

Veo’s API route matters. The Gemini Developer API preview retirement is not a shutdown of Google Cloud’s GA model. Use the exact endpoint’s current documentation rather than mixing a preview model’s duration, price, and extension rules with the Cloud route.

Seedance 2.0 is not the whole Seedance family. Full 2.0 supports 4–15-second base generation; Fast and Mini have different resolution options. Seedance 2.5 documents up to 30 seconds. The earlier claim that nothing could replace a historical 20-second Sora shot on duration alone omitted that newer model. Duration support does not establish equivalent motion or visual quality.

Compare the models by the job

Long shots and a finished sequence

Keep base generation, multi-shot content, extension, and final assembly separate. A clip can contain several camera cuts while still fitting inside one model’s duration ceiling. Extending a source clip is another task; assembling several approved clips in an editor is another workflow again.

Full Seedance 2.0 supports 4–15 seconds per base request; Seedance 2.5 supports 4–30. Cloud Veo 3.1 supports 4/6/8-second base requests. Its extension guide adds seven seconds to an eligible input of up to 30 seconds, reaching up to 37 seconds total. This does not establish a 60- or 148-second extension allowance for that Cloud route.

If you are replacing an old Sora long shot, start with the same framing, action, and reference assets. Compare the completed candidates. This article does not establish a physics winner or promise a replacement will match the historical output.

Native audio and the option to generate silence

Veo 3.1 supports synchronized speech and sound effects. Google Cloud lists video-only and video-with-audio variants, so “audio is always on with no audio-off option” is not true across its routes. Select the intended variant and review speech, sound, and lip-sync in the result.

Seedance 2.0 also documents native audio, enabled by default, plus a provider audio-off control. The inspected ChatCut Seedance route keeps audio enabled. Native provider options and exposed ChatCut controls must be described separately.

References and multiple shots

Seedance 2.0 accepts up to nine image references, three video references, and three audio references within the request’s combined limits. Audio references need an image or video reference; frame mode cannot be mixed with reference mode. Do not add the individual maxima together and assume every combination is valid.

The official Seedance 2.0 guide includes fast cuts and ordered shot prompts within one request. Multi-shot generation is a model capability, not only a manual chaining technique. There is no established twelve-output-shot guarantee. For 2.0, describe “Shot 1,” “Shot 2,” and so on; exact timestamp following is not a guaranteed timing control. Seedance 2.5 documents timestamp following separately in its prompt guide.

Cloud Veo 3.1 supports first-frame, first+last-frame, and subject-reference workflows with mode-specific limits. The model card documents up to three subject images for the full GA model. Do not assume all input modes can be combined in one request.

Cross-shot consistency

Use approved references to guide a recurring character or product, then check identity, costume, labels, and geography across the actual cuts. Neither a reference count nor multi-shot support establishes perfect identity preservation.

A larger reference library can be useful, but select no more than nine images for each Seedance 2.0 call and stay within the combined request limits. Carrying a previous approved clip into a later request is an optional workflow; it is not the only way to ask for several shots.

For a longer project, the AI filmmaking workflow covers reference preparation and assembly.

Cost per approved clip

The original comparison mixed retired Sora tariffs, universal per-second Seedance estimates, and a ten-second Veo request outside its documented base duration choices. It also used an unsupported “industry-typical” reject rate. Those numbers do not provide a valid current cost comparison.

Get a quote for a valid request on each route. Record model ID, provider, resolution, aspect ratio, duration, audio choice, reference-video duration, quote date, and any retry budget. Google Cloud’s fetched pricing page lists audio and video-only variants, but its displayed unit needs confirmation against the applicable SKU before treating amounts as per-second prices.

For ChatCut, credits policy owns current rates. Read the preflight estimate and resulting ledger rather than treating an old three-credit clip example as a universal tariff. No current ChatCut Seedance 4K price is established.

Rebuild a Sora workflow after retirement

For long single shots, evaluate available routes with valid durations, including Seedance 2.5 where a 30-second base request fits the brief. Compare actual motion and continuity; duration alone is not a quality comparison.

For ideation, retain the script, shot list, and approved references. Generate a candidate still if useful, animate it with an available model, then review the result in context. This is a replacement method, not a claim that another model reproduces Sora’s exact visual range.

For Cameo-style identity work, reference-image anchoring is a different workflow. Prepare a character sheet and select up to nine images per Seedance 2.0 call. Review identity across shots; this article does not establish feature parity with Cameo or a universal direct successor.

Where ChatCut fits

ChatCut combines image generation, video generation, motion graphics, and editing in a project. Check the current selector and model matrix before choosing a route.

The inspected tool exposes Seedance variants and Gemini Omni, not a Veo selector. Its Kling new-generation route is suspended. Those integration states do not establish the underlying models’ native capabilities or a provider account’s entitlement. Deployed production was not re-tested for this article.

Generate or import a reference still, check the video settings and estimate, review the returned clip, then place it in the cut. Final export has separate renderer settings; see exporting. A larger exported frame does not establish higher-detail original generation.

For the broader category, see the six-generator workflow guide. For Seedance prompting, see the prompt guide.

FAQ

How do I migrate a Sora 2 workflow after retirement?

The September 24, 2026 shutdown date is already past. Rebuild new-generation dependencies on an available route; keep exported assets you already hold. Confirm that any replacement endpoint is current—Gemini API Veo 3.1 preview also retired on October 22, while Cloud GA is separate.

Does Seedance 2.0 actually replace Sora 2?

It can serve some of the same shot-generation jobs, but it is not established as a drop-in replacement. Full 2.0 supports 4–15-second clips and multiple ordered shots; Seedance 2.5 separately supports up to 30 seconds. Test your action, references, and sound rather than assuming equivalence.

Why isn’t Kling in this comparison?

This article follows the three model names in the query. For a wider set of workflows, see the six-generator guide. Check route availability separately; ChatCut’s inspected Kling new-generation route is currently suspended.

What does ChatCut actually charge to generate one Seedance 2.0 clip?

The estimate depends on the exact model and settings, including duration, resolution, aspect ratio, and reference-video input. Check credits policy and the displayed estimate. The old flat three-credit example is not a current quote for every clip, and access to free editing does not establish that a generation is free.

Try the reference workflow

Pick one shot from a current project. Generate or import a reference still, select an available video route, review its estimate, and generate a candidate clip. Check motion, identity, and sound before assembling the final cut.

Open ChatCut’s AI Video Generator →