AetherCutChangelogHow it worksPricingOpen the editor

Home / Changelog / v1.9.85

v1.9.85

Flagship trio launch — Prompt-to-Cut, AI Reshoot, Cross-Language Lipsync

AetherCut v1.9.85 shipped on Aug 13, 2026.

What's new in v1.9.85?

  • Cross-Language Lipsync (flagship #3) — Whisper transcribe + ElevenLabs Dubbing (voice-preserving translate) + fal.ai sync-1.7.1 lip resync, chained inside a single editor timeline. First editor to combine all three in one click.
  • AI Reshoot (flagship #2) — highlight any clip, describe what you want instead, Sora 2 regenerates a matching segment (aspect ratio, duration, transcript context all auto-inferred). Insert-after is safe default; Replace exists for one-tap Ctrl+Z undo.
  • Prompt-to-Cut (flagship #1) — describe the edit in plain English, AetherCut transcribes every clip, Claude plans the cut, timeline auto-assembles. Progress UI shows transcribe/analyze/apply phases so a 30-60s process doesn't feel broken.
  • Save Transcript byproduct — Prompt-to-Cut generates transcripts on the way to the plan; a new button on the result card lets you download them as .txt with zero extra spend.
  • Marketing counts un-ghosted: 56 → 59 Pro tools, 28 → 31 AI features. Every marketing surface (landing, About, comparison tables, share text) auto-syncs.
  • Strategic update dialog — two-step disclosure explaining the pay-at-export suspension + previewing the new flagship lineup. Once per browser, gated behind bug-disclosure + onboarding so users never see stacked modals.

Getting useful results from the flagship trio on the first pass

The fastest way to understand this release is to treat the three headline features as different entry points into the same workflow rather than as isolated tools. Prompt-to-Cut is the best starting point when you have raw footage and know the outcome you want but do not want to build the sequence by hand. A practical first pass is to be specific about structure, not just topic: instead of asking for “a short product video,” ask for “a 45-second cut with a strong hook in the first 3 seconds, one customer quote, one demo moment, and an ending CTA.” That gives the planner clearer constraints while still leaving room to choose strong moments from the transcript. If the result is close but not quite right, the biggest mistake is re-running with the same vague prompt and hoping for a different arrangement. It works better to tighten one variable at a time, such as pacing, inclusion rules, or the exact moment you want near the opening.

AI Reshoot becomes most useful after that initial assembly, when the edit reveals a gap you did not notice in the raw material. A common example is a sequence that has the right spoken line but weak visuals around it: maybe the speaker says the key benefit clearly, but the original shot has a distracting background, awkward framing, or no product action to support the claim. Highlighting only the problem span and describing the replacement in terms of what should be visible usually works better than trying to redesign the whole scene in one go. The auto-inferred duration and context help the regenerated segment fit, but it still pays to keep your instruction anchored to the surrounding edit. If the clips before and after are calm and static, asking for a frantic camera move can make the inserted shot feel like it belongs to a different project. For that reason, “same framing and pace, but show hands using the product on a desk” is often a stronger prompt than something broader like “make this cinematic.”

Cross-Language Lipsync tends to go most smoothly when you use it on lines that are already carrying the meaning cleanly in the source cut. In practice, that means locking the story first, then translating, instead of dubbing too early and discovering later that you still need to move scenes around. Because the chain handles transcription, voice-preserving dubbing, and lip resync inside the timeline, the main user-side judgment call is editorial: which moments truly need to be spoken on camera, and which can be covered with B-roll or text? If a line is visually front-and-center, lipsync earns its keep. If the speaker is off-screen or the shot is wide enough that mouth detail is minor, you may not need to spend effort there. One easy pitfall is treating every sentence as equally important across languages. A stronger approach is to watch the translated cut for rhythm, not just accuracy, and then use small timeline trims where needed so the dubbed performance lands naturally inside the scene.

There is also a simple order of operations that avoids extra backtracking. Start with Prompt-to-Cut to generate the rough structure and save the transcript while it is already being created. That transcript is useful for quick review, approvals, or checking whether the planner is selecting the right material before you fine-tune visuals. Then use AI Reshoot only where the assembled timeline exposes missing coverage or weak shots, because those problems are much easier to identify in context than in isolated source clips. Finish with Cross-Language Lipsync once the timing and story are mostly stable. Users who reverse that order often create unnecessary cleanup for themselves, especially when a late structural edit forces them to revisit translated or regenerated sections. Taken together, the trio works best when each feature is used for the decision it is actually good at: Prompt-to-Cut for selecting and sequencing, AI Reshoot for repairing visual gaps, and Cross-Language Lipsync for localization after the edit already deserves to travel.

Which features does this release touch?

How do I get this update?

Open the editor and you have it. AetherCut runs in Chrome, Edge, Safari and Firefox. There is nothing to install, and you can edit without an account. Open the editor or see how it works.

What shipped just before and after?

See all 111 releases · Follow the RSS feed