Generative Media Tool is the name Adobe gave, on 8 September 2026, to the thing every editor has been doing badly for two years: leaving the timeline to go and make something, then coming back with a file that does not match the cut. Draw a range on any track, describe what is missing, pick a model, and the Generative Media Tool fills the gap in place. The Verge summarised the intent bluntly — Adobe is trying to make its AI generators idiot-proof.
The Generative Media Tool interface change is real and it is good. What follows is not a complaint about it. It is an audit of the one number the announcement leaves out, because that number is the only part of this release an editor cannot recover from by pressing undo.
Between Adobe’s own launch post, The Verge’s write-up and ProVideo Coalition’s hands-on, three documents describe five models, four generators, four new audio tools and four mask types. Across all three, the count of published credit figures is zero. ProVideo Coalition puts the omission plainly: the cost “shows in the taskbar before you commit,” and “credit usage depends on the model chosen.”
So this piece does the division the announcement skips. Every figure below comes from Adobe’s own generative credits page or from arithmetic performed on it, and the result is a simple one: a five-second generated clip costs 100 credits, the cheapest plan holds 2,000 of them, and the taskbar that tells you the price of this generation never tells you it just spent 5% of your month.
Table of contents
- What the Generative Media Tool Actually Changes in Premiere
- Five Generative Media Tool Models and One Published Rate
- The Credit Arithmetic Behind the Generative Media Tool
- What a Month of the Generative Media Tool Actually Buys
- The Generative Media Tool Click That Returns Sixteen Clips
- The Audio Tools Beside the Generative Media Tool Cost Nothing
- After Effects Gets the Assistant That Writes Expressions
- Color Mode Now Carries Four Mask Types
- Budgeting for the Generative Media Tool on a Real Project
- What the Generative Media Tool Announcement Gets Right
- Frequently Asked Questions About the Generative Media Tool
- References and Further Reading
What the Generative Media Tool Actually Changes in Premiere
The mechanism first, because the workflow claim and the cost claim are separate things and only the first one survives inspection intact.
One range, one taskbar, no app switching
The Generative Media Tool lives in the Premiere toolbar in the same slot as Generative Extend; a long press toggles between the two. Select it, drag a range across an empty gap on a video or audio track, and a generative taskbar appears next to that range. Write a prompt, choose a model, generate. The clip lands in the gap already trimmed to the range you drew.
Premiere reads the sequence settings and applies them, so the returned footage arrives at the project’s resolution and aspect ratio rather than at whatever the model defaults to. That single detail removes the most common reason generated B-roll used to need a round trip through another application.
Reference frames sampled from your own edit
An eyedropper lets you set one or two reference frames by sampling directly from the timeline or the program monitor. The generated footage is then conditioned on shots that already exist in the cut, which is the difference between a plausible clip and a clip that matches the scene either side of it.
This is the part of the Generative Media Tool that genuinely reduces the number of attempts. Every retry costs credits, so anything that raises the hit rate on the first generation is worth more than it looks.
Four generators behind one Generative Media Tool control
The Generative Media Tool is not one generator but four, two of them shipping and two in beta. They share the range-drawing gesture and nothing else — different models, different outputs, different maturity.
| Generator | Model choice | What comes back | Status |
|---|---|---|---|
| Generate Video | 5 models | One clip, sequence-matched | Shipping |
| Generate Sound Effects | Adobe audio model | 4 variations | Shipping |
| Generate Music | Firefly Music | Multiple variations, tempo and loop controls | Beta |
| Generate Soundscape | Automatic, no prompt | 4 tracks x 4 variations = 16 clips | Beta |
Five Generative Media Tool Models and One Published Rate
Here is where the Generative Media Tool stops being a workflow story and starts being a procurement one.
The model list
Inside the Generative Media Tool, Generate Video offers Adobe Firefly plus four partner models: Google Veo, Kling, Runway and Luma. ProVideo Coalition notes the list should be expected to change as new models are added, which is sensible product design and a moving target for anyone trying to budget.
Adobe’s framing is choice — pick the model that fits the shot. That is a real benefit, and it is also the moment the editor is asked to make a commercial decision inside a creative gesture, with only one of the five options carrying a rate anyone can look up in advance.
Why the model choice is the expensive decision
Firefly is the Generative Media Tool default, and its consumption is documented on Adobe’s public credits page in the form of plan examples. The four partner models are premium: Adobe states that credit usage depends on the model selected, the output and the file size, without publishing a rate card for any of them.
The practical consequence is that four of the five entries in the Generative Media Tool dropdown have a price you can only discover by configuring the generation and reading the taskbar. That is not hidden pricing — it is just-in-time pricing, and it is fine for one clip and useless for a month’s plan.
The credit line in the Generative Media Tool taskbar
To Adobe’s credit, the cost is shown before you commit, which is more than most generative tools inside professional software do. The number is accurate, immediate and specific to the model and settings you have chosen.
It is also scoped to a single generation. Nothing in the Generative Media Tool taskbar tells you what fraction of the monthly allowance this generation represents, and that fraction is the figure that determines whether the feature is usable on a real project.
The Credit Arithmetic Behind the Generative Media Tool
Adobe does publish enough to compute the Generative Media Tool’s Firefly rate exactly. It just publishes it two clicks away from the announcement, as marketing examples rather than as a rate.
The four packs, divided out
Adobe’s generative credits page lists four monthly credit packs and, for each, an example of how much video it produces. Divide the second by the third and the same number falls out four times.
| Pack | Price per month | Adobe’s stated output | Credits per clip | Cost per clip | Cost per 1,000 credits |
|---|---|---|---|---|---|
| 2,000 credits | US$9.99 | up to 20 five-second videos | 100 | $0.4995 | $4.995 |
| 7,000 credits | US$29.99 | up to 70 five-second videos | 100 | $0.4284 | $4.284 |
| 10,000 credits | US$49.99 | up to 100 five-second videos | 100 | $0.4999 | $4.999 |
| 50,000 credits | US$199.99 | up to 500 five-second videos | 100 | $0.4000 | $4.000 |
One hundred credits per five-second clip, confirmed four ways
2,000 divided by 20 is 100. 7,000 divided by 70 is 100. 10,000 divided by 100 is 100. 50,000 divided by 500 is 100. Four independent pack sizes agree on the same Generative Media Tool rate, which is strong evidence it is the real Firefly rate rather than a rounded marketing figure.
One hundred credits for five seconds is 20 credits per second of generated video. That is the number the Generative Media Tool spends every time you draw a range and press generate, and it is the number the announcement never states.
The pack that costs more per credit than the pack below it
Dividing price by credits produces the last column above, and it is not monotonic. The 7,000-credit pack costs $4.284 per thousand credits. The larger 10,000-credit pack costs $4.999 per thousand — 16.7% more per credit for buying more of them.
Two 7,000-credit packs cost $59.98 and yield 14,000 credits. One 10,000-credit pack costs $49.99 and yields 10,000. Spending 20% more money buys 40% more credits. Anyone standardising a team on the $49.99 tier because it is the obvious middle option is choosing the single worst rate Adobe sells.
What a Month of the Generative Media Tool Actually Buys
Generative Media Tool credits are an abstraction. Seconds of finished footage are not, and the conversion is a single multiplication.
Minutes of generated video per pack
Five seconds per clip, multiplied by the clip counts Adobe publishes, gives the monthly ceiling for each pack. The results are smaller than the credit numbers make them sound.
The largest pack Adobe sells is US$199.99 a month, or $2,399.88 a year, and it buys 41 minutes 40 seconds of generated video per month. Over twelve months that is 500 minutes — eight hours and twenty minutes of footage, before a single retry.
Video costs four times what translation costs
The same credits page gives audio translation allowances: 6, 23, 33 and 166 minutes for the four packs. Divide credits by minutes and all four land on 300 credits per minute once you allow for Adobe rounding down, which is 5 credits per second of media.
Generated video is 20 credits per second. Translation is 5. The Generative Media Tool spends credits four times faster per second of media than the translation feature does, which is worth knowing before you decide which AI feature your team standardises on.
The percentage the Generative Media Tool never shows
On the 2,000-credit plan, one five-second Generative Media Tool generation is 5% of the month. A thirty-second social spot generated end to end is 600 credits, or 30% of the month, and that assumes every clip lands first time.
The taskbar will correctly report 100 credits. It will not report that this is a twentieth of everything you have. That gap between an accurate per-generation price and an invisible monthly budget is the one place the idiot-proofing stops.
The Generative Media Tool Click That Returns Sixteen Clips
The Generative Media Tool beta generator with the largest output is also the one with no prompt and no published rate, which makes it the sharpest illustration of the same problem.
Fifteen seconds in, sixteen files out
Inside the Generative Media Tool, Generate Soundscape skips the prompt entirely. It runs computer vision over up to 15 seconds of the video above your selected range and returns four pre-mixed tracks — two ambience, two sound effects — each with four variations, timed to the action. That is sixteen audio assets from a single gesture.
Adobe frames it as a starting point, a sketch to build on rather than a finished mix. ProVideo Coalition’s hands-on is more measured, reporting mixed results: some tracks usable, others out of sync and off the mark.
The generator with no prompt and no rate
Generate Sound Effects returns four variations per prompt. Generate Music returns multiple variations with tempo and looping controls. Generate Soundscape returns sixteen. None of the three has a published credit rate anywhere in the announcement.
For a shipping video generator, just-in-time pricing in the taskbar is a reasonable design. For a beta generator that produces sixteen assets from one unprompted click, the absence of a rate in the documentation is a larger gap, because the editor has no prior for what “one click” ought to cost.
Why the count matters more than the quality
Sixteen clips is not a quality problem, it is an audition problem. Every returned variation has to be listened to, and the time cost of that is borne by the editor whether or not the credits were well spent.
This is the recurring shape of generative AI inside professional tools: generation gets cheaper and faster, evaluation does not. Our AI Models, Tools and Releases hub tracks the same pattern across the wider tooling market.
The Audio Tools Beside the Generative Media Tool Cost Nothing
Not everything in this release runs through the Generative Media Tool, and the non-generative half is the part most editors will use most days. It costs no credits at all.
Enhance Audio splits one clip into four components
Enhance Audio takes a mixed clip and separates it into dialogue, music, ambience and sound effects, with an independent slider for each in the Properties panel. You can pull the music down for clearance reasons, reduce room tone without stripping it, or mute dialogue entirely.
A companion feature, Separate Sound Source, promotes those separated components onto their own physical tracks so they can be handed to a mixer or processed independently. One clip in, up to four tracks out.
Separate Crosstalk turns two speakers into three tracks
Separate Crosstalk splits a recording of two overlapping speakers into two independent tracks, plus a third holding everything else. It was first shown as a sneak at Adobe MAX and has now reached public beta.
For anyone who has ever tried to cut a two-hander interview recorded on a single shotgun microphone, this is a bigger practical win than anything in the Generative Media Tool, and it does not consume a credit.
Dynamic Auto Ducking regenerates on every edit
Dynamic Auto Ducking writes the keyframes that balance dialogue against music, and rewrites them whenever clips move in the timeline. The reason ducking is so often skipped is not that the keyframes are hard, it is that they have to be redone after every trim.
| Tool | Input | Output | Credits |
|---|---|---|---|
| Enhance Audio | One mixed clip | 4 independent component sliders | None stated |
| Separate Sound Source | One mixed clip | Up to 4 physical tracks | None stated |
| Separate Crosstalk | 2 overlapping speakers | 3 tracks | None stated |
| Dynamic Auto Ducking | Dialogue plus music | Keyframes, regenerated on edit | None stated |
After Effects Gets the Assistant That Writes Expressions
The After Effects half of the announcement is a different bet from the Generative Media Tool: not generating media, but operating the application on your behalf.
Six pre-comps from a single prompt
AI Assistant is now in public beta in After Effects, having run in Premiere beta for several weeks. In the press demo, Adobe’s Jason Druss built a 1990s stop-motion collage animation from a Photoshop file, with the assistant writing expression rigs for six photo-frame pre-comps and adding slider controls for wiggle amount, frame rate, stack spread, path messiness and several other parameters — from one prompt.
It works across an entire project rather than the composition you have open, which is what makes the organisational use cases viable at all.
Every action undoable because it uses native tools
The assistant drives native After Effects operations rather than a parallel mechanism, so everything it does is undoable. It shows its tool calls, explains its reasoning, presents parameters in tables and asks clarifying questions before acting.
That design choice is the actual idiot-proofing in this release. A generated clip costs credits whether or not you keep it; an assistant action costs nothing to reverse. The asymmetry is worth noticing when deciding which of these features to hand to a junior editor.
Project organisation is the sleeper feature
The demo-friendly capability is expression writing. The one that saves the most hours is being handed an inherited project with thousands of layers and having the assistant propose a structure and carry it out after you approve the plan.
This is workflow automation applied to a creative tool rather than a back office, and the economics are the same: the value sits in the tedious multi-step work, not the headline capability.
Color Mode Now Carries Four Mask Types
Color Mode remains in beta and keeps absorbing the requests its beta users file, which is the most encouraging signal in the release.
HSL and Luminance join Object and Shape
Two new mask types arrive: HSL Masks and Luminance Masks, isolating ranges of hue, saturation or brightness. With the existing Object and Shape Masks that makes four mask types, each attachable to any clip operation.
The use case Adobe reaches for is correcting the colour of a face without touching the rest of the frame, which is exactly the kind of selective work that used to force a round trip to a dedicated grading application.
Auto Color across three scopes
Auto Color gives a one-click starting point for colour temperature and exposure, and it runs at three scopes: a single clip, a hand-picked group, or the entire sequence. That range comes from how Color Mode’s operations system is built rather than from three separate features.
Two new Style Modules
Vignette adds corner darkness, photorealistic blur and chromatic aberration. Midtones adds bidirectional curve control for editors who wanted finer tonal adjustment. Both were requested by Color Mode users, and drag-and-drop support for masks rounds out the update.
Alongside all of this, Premiere collected an Engineering, Science and Technology Emmy Award at the 2026 Emmys, recognising 35 years of the application — announced by the Television Academy five days before this release.
Budgeting for the Generative Media Tool on a Real Project
The Generative Media Tool arithmetic above resolves into three questions worth answering before enabling any of this across a team.
Count Generative Media Tool clips, not seats
Licence cost is the wrong unit. The Generative Media Tool consumes credits per generation, and a team of four editors sharing a 7,000-credit pack has 70 five-second clips between them for the month, minus every retry.
Estimate the number of gaps a typical project needs filled, multiply by an honest retry factor, and multiply by 100 credits. That figure, not the seat count, tells you which pack to buy.
Price the partner models before standardising
Firefly’s rate is computable. Veo, Kling, Runway and Luma are not, from public documentation. If a house style settles on a partner model, run a handful of test generations and read the taskbar figures before committing a project to it.
Separate the credit features from the free ones
Enhance Audio, Separate Crosstalk, Dynamic Auto Ducking, Color Mode and the After Effects assistant carry no stated credit cost. The four generators inside the Generative Media Tool do. Rolling out the first group is a training question; rolling out the second is a budget question, and treating them as one release is how the surprise arrives at the end of the month.
For teams formalising this, the same discipline that applies to any AI strategy applies here: measure consumption on a pilot before writing the policy.
What the Generative Media Tool Announcement Gets Right
It would be unfair to end the Generative Media Tool audit on that omission. Adobe’s own framing cites its 2026 Creators’ Toolkit Report, which found that nearly nine in ten creators say creative AI tools accelerate the growth of their business or audience, alongside an overwhelming majority saying final creative decisions must stay with them.
The Generative Media Tool is built to that brief. Everything it returns is a fully editable clip on your timeline rather than a locked render, the model choice stays with the editor, and the credit cost is disclosed before the generation runs. Frame.io Mounted Storage reaching all accounts, and the redesigned Properties and Effects panels, are unglamorous fixes that will save more time than the headline feature.
The idiot-proofing worked on the craft decision. It has not yet reached the budget decision, and one line in the taskbar showing the monthly balance would close that gap entirely.
Frequently Asked Questions About the Generative Media Tool
What does the Generative Media Tool actually do in Premiere?
It lets you draw a range on a Premiere timeline track, describe what is missing, choose a model and generate video, sound effects, music or a soundscape directly into that gap, without leaving the application.
Which models can I choose from?
Generate Video offers Adobe Firefly, Google Veo, Kling, Runway and Luma. Adobe has said the list will change as models are added.
How much does one generation cost?
Adobe’s public credits page implies 100 credits for a five-second Firefly video, consistent across all four credit packs. Partner model costs are not published and are shown in the taskbar before you commit.
Which credit pack offers the best value?
The 50,000-credit pack at US$199.99 is cheapest per credit at $4.000 per thousand. The 10,000-credit pack at US$49.99 is the most expensive at $4.999 per thousand — worse than the 7,000-credit pack below it.
Do the new audio tools consume credits?
Enhance Audio, Separate Sound Source, Separate Crosstalk and Dynamic Auto Ducking have no stated credit cost in any of the announcement documents. They are in public beta in Premiere.
Is the Generative Media Tool available now, and which parts are beta?
Generate Video and Generate Sound Effects are shipping with the IBC 2026 release. Generate Music and Generate Soundscape are in beta, as are the audio repair tools, Color Mode and the After Effects AI Assistant.
References and Further Reading
The Verge: Adobe is trying to make its AI generators idiot-proof in Premiere
ProVideo Coalition: Adobe’s IBC release for Premiere and After Effects
Adobe: Add Generative Credits to keep creating
Adobe Help: Generative Extend in Premiere
Television Academy: Engineering, Science and Technology Emmy Awards
More AI coverage: explore Progressive Robot's AI Models, Tools & Releases hub — hands-on reviews, setup guides and benchmarks in one place.