How to Turn a Photo Into an AI Video, One Scene at a Time

Begin with one photo and a short action. You can animate a portrait's expression, introduce movement into scenery or create a restrained product shot. A short clip lets you identify what changed before building a longer project around it.

This tutorial follows an existing Dreamina portrait example through a saved native file. The portrait depicts a fictional AI-generated adult, not a real person's identity.

1. Choose a photo that supports the movement

Select the sharpest useful source, rather than the largest file regardless of content. The important subject should be visible, and the framing should leave room for the movement you intend.

For a small head turn, choose a clearly visible face with enough space around the head and shoulders. For walking, use a fuller-body composition. For a slow product reveal, keep the product's top, bottom and important label inside the intended frame.

The example began with this 2048 × 2048 synthetic portrait:

The short curly hair, blue shirt and plain gray background provide recognizable details to compare with the moving result. This is an AI-generated demonstration subject.

Inspect existing problems before uploading. Motion generation can carry forward a malformed hand, unclear eye or distorted label. If the photograph already needs repair, correct that image or choose another source first.

Match the image shape to the intended video where possible. A square portrait is a straightforward start for a square clip; squeezing it into a narrow vertical crop can create a different framing problem. The input-photo guide explains how to choose a source for different actions.

2. Open an image-capable video workflow

In Dreamina AI Video, choose an available image/reference-capable video workflow and model. Its Seedance 2.5 guide describes adding reference inputs, directing the scene and setting the output.

Upload the intended photo and check that it appears in the input area. Remove references left from a previous task. For this first project, one image is enough; do not add unrelated material simply because the interface accepts more inputs.

A reference-led mode and a first-frame mode can use an image differently. Inspect the option you select rather than assuming every uploaded reference is locked as the exact opening frame.

Choose the model available in your account and set a short duration for the first scene. In Dreamina, a clear reference image and one small action provide a focused starting point for exploring motion.

3. Choose a short duration and suitable format

Set the intended aspect ratio, duration and available resolution. Read the quote after making those selections, since a change of model or output can change the cost.

The recorded example used:

Those are the submitted settings, not a general price or a guarantee that every square video has the same dimensions. The saved file was 960 × 960. Inspect the actual export rather than interpreting a resolution label as a universal width and height.

A short first attempt gives you enough room for a modest expression or camera move. If you ask for a long sequence of actions, the model may compress or omit parts. Begin with the smallest motion that answers your purpose.

4. Describe what should move

Let the image provide the visible subject and use the prompt to direct the action. Runway's image-to-video prompting guide makes this distinction explicit: the image establishes visual information, while the text describes movement and progression.

The actual portrait instruction was:

The same fictional adult blinks once, turns her head slightly to her left and returns her gaze to the camera with a subtle relaxed smile. Preserve her exact facial proportions, short curly hair, blue shirt and gray background. Natural restrained movement, soft portrait light, no speech, no new objects.

The clip must still be reviewed.

For your own photo, replace the subject and action with what the source can support. You might ask for a gentle ear movement in a pet photo or slow drifting clouds in a landscape. Avoid introducing several new subjects, a dramatic camera orbit and complex hand contact in the same first attempt.

Decide whether the camera should move. A small subject action with a fixed camera is easier to evaluate than an output where both move unpredictably. If the tool provides a dedicated camera control, check how it interacts with the reference mode instead of duplicating contradictory instructions.

5. Generate and compare the result with the source

Submit once the photo, model, settings and quote are correct. Keep the prompt and original input so a revision can address a specific failure.

Here is the native portrait result:

The file is approximately 5.055 seconds at 960 × 960 and has an audio track. Recorded half-second frame samples show blinking, a head turn and a developing smile, with the hair, shirt and face broadly recognizable.

Compare the result with the job you requested. Did the subject move in the intended direction? Did the expression develop naturally? Did the clip return to a useful ending? Check the eyes, mouth, hairline, clothing and background separately.

For your own acceptance, watch the full video at its intended viewing size. A thumbnail may look good even when the face changes briefly between the sampled moments.

The example requested no speech. It should not be treated as a talking-photo demonstration just because the mouth changes with the expression or the file contains a sound track. A speaking portrait needs a specific performance or lip-sync workflow and an independent speech check.

6. Correct the cause of the failure

Change one meaningful part of the setup so you can tell whether it helped.

Use a clear source image with enough detail for the chosen mode. If an upload is too small, replace it with the highest-quality original rather than enlarging a thumbnail; the input panel shows the supported requirements.

If generation stops before a result appears, keep the source and prompt, check the message and retry only after resolving the input or account issue. This preserves the creative direction without starting the project over.

7. Download and verify the saved video

When you accept a result, use the available native Download option. The portrait example was saved through Dreamina's native download control.

Open the file after saving. Check its full duration, dimensions, sound, visible marking and beginning and ending. A result visible in the browser is not yet a usable local export. Keep the original image and the accepted clip together.

If you need music or narration, add it deliberately afterward or select a model that includes the required audio. The free photo-to-video sound guide separates native audio from later editing and account eligibility.

One downloaded clip completes this tutorial's main task. To build several shots, titles and sound into a larger finished piece, continue with the complete AI video project workflow.

Alternative: use Firefly's image workflow

Adobe's current image-to-video help provides a concrete alternative:

1
1. Open Video, then Generate video, and select Firefly Video.
2
2. Choose the available output settings and add the image under Frame → First.
3
3. Adjust the crop if needed, write the motion prompt and generate.
4
4. Review the clip and download it from the result or history.

The controls depend on the input. Adding a first frame disables some style and composition controls; adding an end frame also turns off the Motion options. A last frame can guide an ending, but should not be assumed to work with every camera option simultaneously.

Firefly also offers partner models. Their input and audio capabilities differ, so follow the selected model's instructions rather than transferring every Firefly Video control to it. Its product overview describes free daily generation with the Adobe model; check the actual account and available configuration before planning repeated work.

Plan for attempts, not just the clip length

Free access and paid upgrade: Free accounts in the US receive 120 credits per day; daily allowances vary in other regions. Start developing images and short videos with Seedance 2.0 for video and Seedream 5.0 Lite for images. For regular production, Dreamina Basic is USD 15 per month on monthly billing, with 1,575 monthly credits. Paid plans add watermark-free downloads and a faster generation queue. Credit use varies by model and settings.

If a revised action needs another generation, it can use another portion of the allowance. Keep a small first project and decide what would make the result acceptable before submitting. When free access is the deciding condition, check the image-to-video allowance guide first.

Begin with a photo whose subject you can recognize and one movement you can judge. Upload, direct, inspect and save that clip; add a longer story only after the single-image workflow produces material you can use.