Firefly first and last frame controls determine which camera options remain available. For the Firefly Video model, adding a first image removes several selectors; adding a last image also turns off Motion. Keep both frames when the ending composition matters, or use a first frame alone when a selectable Motion preset is the priority.
This guide covers the Firefly Video model’s web workflow. Partner models inside Firefly may offer different controls.
Choose the constraint that matters most
Suppose you want a short video of a paper lantern above a wooden table. You already have a strong starting image. There are two different briefs: show that composition with a gentle camera move, or travel between it and a second, carefully arranged ending image. Decide which brief you mean before adding both references.
Adobe’s image-to-video instructions say that a first-frame image disables several selectors, including shot size, camera angle, style, composition reference and transparent background. With only the first frame, Motion options remain available; adding a last frame disables Motion. A text prompt is required with the first frame alone, while first and last frames make it optional.
| Your priority | Initial setup to consider | What to inspect before generating |
|---|---|---|
| Preserve an opening composition | Upload a first frame | Crop, subject placement and the image itself |
| Explore a camera movement | Start with the first frame alone | Available Motion selection and compatible prompt |
| Reach a planned final composition | Supply first and last frames | Whether the transition is visually plausible |
| Change the starting viewpoint substantially | Rework the source image first | Whether the new source actually shows the desired view |
When a selector becomes unavailable after adding an image, revisit the inputs before treating the missing control as an account problem.
Adobe Firefly image-to-video from upload to download
Use an image you created or have permission to upload and animate. For the lantern example, a photograph of your own table and lantern is enough to plan the shot. Adobe documents this sequence:
- In Firefly, choose Video, then Generate video. Under General settings, select Firefly Video, a resolution and an aspect ratio.
- Under Frame, choose First and upload your image. Hover over its thumbnail for Crop Image if the framing needs adjustment.
- Keep this first attempt to one image. Choose Camera > Motion, then enter a prompt. For a still camera, choose Static; for the example below, try Zoom in.
- Check the credit budget, select Generate, then review the result. Download it from the history strip, or reopen it under Files > Generation History.
Feature access can vary by region and account. The steps here describe Adobe’s documentation, not a signed-in generation test.
For the lantern brief, inspect three things before writing a prompt: whether the lantern is fully inside the crop, whether its support is visible, and whether enough background exists for the intended movement. A tightly cropped image is a poor starting point for a wide reveal because much of the requested scene is absent.
Write the action separately from the scene description. An original example is: “The paper lantern sways gently above the table while the camera slowly moves closer. Warm light remains steady.” It names one subject action, one camera action and one lighting condition that should remain stable.
Adobe’s video prompting guidance recommends clear action, camera direction and timing. Use that structure to resolve ambiguity rather than add decorative language. If a fixed composition is essential, avoid simultaneously requesting a substantial orbit around it.
Add the last frame only when the ending earns its place
An end image is useful when the final composition is part of the brief. It is less useful when it is merely another attractive image. Two individually good photographs can imply incompatible lens positions, object sizes or backgrounds. The generator must bridge those differences, and the transition may be the very thing you did not intend.
Before adding the second image, compare the pair side by side. When the pair is ready, upload the ending image under Frame > Last. Mark which changes are intentional. If the lantern ends lower in the frame, is that because it moved, the camera moved, or the crop changed? If you cannot answer, simplify the pair before spending credits.
| Difference between frames | Editorial question | Possible preparation |
|---|---|---|
| Subject becomes much larger | Is a move toward the subject intended? | Match scale if the desired action is only a sway |
| Table changes angle | Is a viewpoint transition part of the shot? | Prepare frames with compatible perspective |
| Background changes completely | Is the brief actually a scene transition? | Treat it as an intentional transition or separate clips |
| Text differs between images | Must exact lettering remain readable? | Keep critical text as a separate editing layer |
These are review questions, not undocumented Firefly settings. The benefit is deciding where the intended motion comes from before interpreting a strange result as a prompting failure.
Budget credits around decisions you still need to make
Adobe distinguishes having generative credits from having access to a premium feature. Its credits overview says allocations depend on the plan, consumption depends on the feature, and premium features require corresponding access. A balance alone does not establish that a particular video configuration is available.
As of October 1, 2026, Adobe’s generative credits FAQ lists these metered Generate Video rates for the Firefly model at 24 FPS:
| Resolution | Credits per second | Five charged seconds |
|---|---|---|
| 540p | 20 | 100 credits |
| 720p | 50 | 250 credits |
| 1080p | 100 | 500 credits |
Multiply the rate by the duration, then by the number of attempts. If each attempt is five seconds at 720p, three alternatives plus one correction reserve means 50 × 5 × 4 = 1,000 credits. Five seconds is the input to this worked example; use the duration offered by your selected configuration. This buys four attempts, not four approved clips.
Compare the calculation with your account’s current offer and any charge shown before generating. Model, resolution and plan offers matter; these listed rates are not a universal bill for every Firefly user. If the account shows a charge of C credits per attempt, the same four-attempt reserve is 4 × C. Check your remaining balance through your Adobe Account profile menu.
Stop the experiment when the remaining uncertainty belongs in the source image or edit. If the lantern’s printed label must be exact, repeatedly regenerating the entire shot to repair one letter is an expensive way to solve a typography problem. Keep the label treatment separate or use a production method appropriate to that requirement.
Review the motion before assembling the final edit
Use a simple acceptance sheet with four rows: opening matches the intended composition, principal action is readable, object geometry remains acceptable, ending works in the intended edit. Give each row a pass or a specific correction. Avoid replacing these decisions with a single vague quality score.
Then review the clip in context. A subtle sway that looks attractive alone may be too busy behind a title. A compelling close-up may leave no room for the next shot’s transition. Crop and pacing are part of usability, not merely finishing touches.
Once the visual clip is usable, assemble it with titles and sound in your editor. Choosing an image reference does not specify a soundtrack or complete that edit. Keep the exported clip as one production asset alongside the other elements.
Retain the source pair, exact prompt, chosen model and any recorded charge with the exported file. If you later change the ending, you can identify whether the revision concerns reference preparation, generation or editing rather than restarting all three.
Firefly first and last frame controls: a quick decision
| Requirement | What to do before another generation |
|---|---|
| The clip must finish on a particular composition | Keep both frames; use their relationship and an optional prompt to explain the transition. |
| The ending is flexible but a Motion preset matters | Remove the last frame and choose the available preset with the first frame. |
| The starting view itself is wrong | Edit or replace the first image; a disabled Camera angle selector is not an instruction to buy more credits. |
This follows the documented Firefly Video controls. Another model inside the same application can expose different options.
Questions that prevent wasted attempts
Why did Motion disappear after I added another image?
Adding an end frame turns off that selector; it does not mean the generated scene must be motionless. Keep both frames when the ending matters. Remove the end frame when choosing a Motion preset is the priority.
Is the text prompt unnecessary when I use two images?
It is optional with both frames. For the lantern example, use it to explain whether the apparent change in size comes from a camera move or the object moving. A short, clear direction is more useful than adding unrelated scene details.
Does this workflow also describe every model inside Firefly?
No. This guide concerns the named Firefly Video model and the cited instructions. Check the selected model’s own controls and credit charge when using another model. A shared application does not make their limits interchangeable.