The first wave of AI creative was easy to demonstrate: generate a face, animate a scene, produce another version. The visible leap happened in the asset itself, and that deserved the attention it received.
What I’m noticing with Claude Fable 5.1 and OpenAI’s GPT-6 Astra goes further. I can increasingly edit pictures and videos, create layers, and work through more of the creative process with AI.
That changes how I work. AI is becoming useful throughout production, including the decisions and revisions between the first idea and the finished piece. The gains extend well beyond assembling generated parts.
Our AI performance creative workflow covers direction, image generation, and animation. Here, I want to examine the editing controls that let a team develop those assets into a finished ad.
Where is the next opportunity in AI creative?
The next opportunity is working through more of the creative process with AI, from generation through editing and revision. Layers and selective changes give teams more control over how an asset develops.
The practical value is keeping the parts that work while continuing to shape the finished ad around the original idea.
Creative production involves returning to earlier decisions. You see an image in context, revise the composition, watch the video, and reconsider the opening. The first output rarely settles every decision.
When a workflow supports those changes, I can stay with the idea longer. I can develop the work through successive edits instead of treating each output as a finished piece to accept or discard.
The editing side gives me more room to explore an idea before approving it. That flexibility matters alongside speed, especially when the first version suggests a better creative direction.
What do Fable 5.1 and GPT-6 Astra change?
In my workflow, the change is a greater ability to create and edit within the same creative process. That includes working with layers and carrying feedback into further revisions.
This is an observation about the model-and-tool setup I use. The available editing controls still depend on the applications and assets involved.
According to Anthropic’s Fable 5.1 announcement, its improvements include writing and extended agentic work. Customer evaluations describe clearer writing and closer adherence to guidance, both relevant to handling creative feedback.
According to OpenAI’s GPT-6 Astra model documentation, its intended uses include reasoning, coding, computer use, research, and document creation. Those capabilities can support production work when suitable tools are connected.
Neither description makes the reasoning model identical to the image generator, video generator, or editing application. The complete setup matters: which files it can inspect, which changes it can execute, and how you review the result.
I would evaluate that setup across a complete creative task. Start with a brief, make a first version, request specific changes, and inspect both the editable work and the final export.
Why do layers matter for AI creative editing?
Layers make separate parts of a composition easier to change without rebuilding the whole piece. In an editing environment that supports them, the team can preserve approved elements while revising others.
That makes feedback more actionable. You can identify the element that needs work and check whether the revision changed only what you intended.
Consider a hypothetical product ad with a background, a product image, and an offer overlay. The offer needs more emphasis, but the approved product image should stay exactly as it is.
A layer-based workflow gives the team a specific place to make that change. The test is whether those elements remain separately editable and whether the exported ad preserves the approved composition.
The same principle applies to video work, although the available controls differ by tool. Test the specific changes your team needs rather than treating every editing environment as interchangeable.
| Production task | Change to request | Decision to protect |
|---|---|---|
| Layer-based composition | Adjust the offer overlay | Approved product image |
| Image revision | Change one selected element | Surrounding composition |
| Video revision | Tighten the opening sequence | Product demonstration and offer |
| Caption correction | Fix wording and timing | Meaning and legibility |
| Version preparation | Adapt the approved master | CTA and required disclosures |
Choose one task your team handles often and document how it works today. That gives you a specific comparison when you introduce a new model or editing setup.
Inspect the revised project as well as the export. Check whether someone can make another change afterward. An editable handoff should support the next revision as well as the current delivery.
What should a good revision brief contain?
A good revision brief identifies the source project, requested change, protected elements, and acceptance criteria. It should specify which layers or other elements need to remain editable.
That gives both a human editor and an AI-assisted workflow a clear boundary. Without it, a small revision can become an unnecessary rebuild.
Be specific about the instruction. “Make it punchier” leaves the editor guessing. “Remove the repeated introduction and begin with the product demonstration” identifies an observable change.
Name the approved master and keep a copy. Identify the layer or sequence to change and the elements to preserve. Specify which editable project files you expect back.
The final review belongs to someone who understands the product and the audience. Technical correctness does not tell you whether the result feels convincing or delivers the intended message.
For campaigns already producing useful results, our guide to iterating on winning ad creative covers the testing decision. The editing brief translates that decision into a controlled production task.
How do you measure whether editing got faster?
Measure the full interval from the creative brief to an approved export, alongside the human effort required. Record where generation ends and editing begins so you can identify which stages improved.
Then measure campaign performance separately. Production efficiency and advertising effectiveness answer different questions, and both matter to the business.
| Measure | What to record | What it reveals |
|---|---|---|
| Time to approved ad | Creative brief to accepted export | End-to-end turnaround |
| Human production time | Editing, review, and repairs | Actual labor requirement |
| Revision rounds | Returns before acceptance | Instruction and output quality |
| Rework | Unrequested changes corrected | Cost of unreliable execution |
| Campaign outcome | Qualified conversions and cost | Commercial value of the creative |
Use comparable briefs when evaluating a new setup. Record the model, tools, source assets, and reviewer so you can explain what changed between attempts.
A workflow that saves editing time but doubles review time may offer little practical improvement. Our AI creative cost guide explains why pricing needs to account for the work surrounding generation.
What should you ask an AI creative partner to show?
Ask to see an idea move from brief through creation, editing, and approval. Request the first version, a specific revision, the editable project, and the final export.
That demonstration reveals how the team handles constraints and corrections. It is more useful for evaluating delivery than a gallery of selected outputs alone.
Have them explain what the model did, what the editing tools did, and what a person decided. A useful answer should make responsibilities clearer and show where human judgment remains necessary.
The commercial opportunity is greater control across production and a shorter path from idea to live test. That is the standard I would use when evaluating an AI performance creative partner.
If production is slowing down your testing, bring a recent brief and its revision history. Book a Free Strategy Call to discuss where the work gets stuck and what to change first.
Like this? Get the next one.
Short emails. New posts as they ship.
