How to Make Your First AI Video From Text for Free

Start with Dreamina to turn a written idea into an original scene using Seedance 2.0. US free accounts receive 120 credits per day, while daily allowances vary in other regions. For a script spoken by a presenter, HeyGen offers a separate free allowance; eligible Google Flow accounts also receive practice credits.

Choose what “from text” should produce

A finished explainer does not need every frame to be newly generated. Conversely, a beautiful six-second clip does not automatically include narration, captions or a complete story.

For a first project, choose one result: a single visible action or one short explanation. You can build a longer piece after you know the generation and download conditions.

Start your free Dreamina video with Seedance 2.0

Log in to Dreamina, open AI Video and choose Seedance 2.0. Enter one short scene description, select the format and duration, review the displayed credit cost, then click Generate. Preview the completed video and click Download to save the MP4. Your free daily credits provide a starting budget for the scene and further variations.

Generate a short first take, watch it through, and refine the subject or movement before making the next variation. You can also prepare the visual direction with Seedream 5.0 Lite, which is available through the free image workflow.

For example, turn “a robot makes a delivery” into a single shot: “A small courier robot stands beside a parcel, raises one hand and waves.” Once that shot works for the story, add the next scene.

Use the scene-writing method below with a model included in your account. For more detail on planning the allowance, see the free-credit guide.

Route 1: turn a short script into a free presenter video

This is the clearer route when the important result is someone speaking your words, such as a small tutorial or explanation. It does not produce an arbitrary cinematic scene from the script.

HeyGen’s Free plan includes three watermarked videos per month, up to one minute each, with stock presenters and video-creation tools. It is a useful starting point for a short scripted explanation.

Prepare a script with one job

Keep the message short enough to understand aloud. Instead of “explain everything about growing plants,” answer one question, such as how to tell whether a seedling needs watering.

Here is an illustrative script you can adapt:

Before watering a seedling, check the soil. Touch the surface gently. If it still feels damp, wait and check again later. If it feels dry, add a small amount of water around the soil, then let any excess drain. Keep an eye on the plant and adjust your routine as it grows.

This is writing for a sample video, not the script behind a demonstrated output. Read it aloud and shorten anything that sounds crowded. Do not fill the one-minute maximum simply because it is available.

Create and export

HeyGen's current beginner guide describes these two starts: Video Agent for a described video, or AI Studio for scene-by-scene control.

1
1. Create a Free account and start a project. Use a stock presenter for a simple first attempt.
2
2. Add your script, or describe the desired short video in Video Agent.
3
3. Select an available voice and the intended format.
4
4. Review scene order, visible text and timing in the preview.
5
5. Generate the final video, then review the rendered speech and lip movement.
6
6. Download the MP4. Keep the free watermark in place.

Check difficult words in the finished render, especially names, product terms and abbreviations. A preview is useful for layout, but it is not the final speaking performance. If you need a correction, confirm how another generation affects the remaining free allowance before submitting it.

Choose this route when a watermarked, short presenter video satisfies the task. A free output can be useful for learning or reviewing an idea without being the file you ultimately want to publish.

Route 2: create an original scene with an available Flow allowance

Google's Flow credit help says non-subscribers receive 50 daily credits, without daily rollover. Its listed non-Ultra rates include 10 credits per Veo 3.1 Lite generation and 20 per Fast generation. Costs are per generation; one request can create more than one video.

Open Flow in a supported region to view the models and allowance available to your account. Google's getting-started guide explains its subscription and qualifying Workspace entry routes.

If you have the non-subscriber allowance and model access, 50 credits can cover up to five ten-credit attempts when used only for that operation. That is an arithmetic ceiling, not five guaranteed usable clips. Non-subscribers may also be unable to generate during the documented peak period of roughly 14:00–17:00 UTC, and free 1080p upscaling is not included.

Set up one inexpensive attempt

In a Flow project, its standard creation controls let you select Video, then the model, aspect ratio, generation length and number of outputs. Do not leave the project in an image model and expect a moving clip.

For an initial scene, choose one output and a short duration offered by the selected model. Read the current quote after changing settings. A reference, feature or incompatible setting can change the model that is used.

You could adapt this new example prompt:

A small coral paper boat drifts slowly from left to right across calm dark-blue water. One soft warm reflection moves with the ripples. Locked wide camera, one continuous shot, no people, no lettering.

Give the model one visible action. This makes it easier to tell whether a failed result comes from the boat's shape, the direction of travel or the camera movement. Several unrelated events in a short clip make both generation and review harder.

Generate only after the account, model and quote satisfy the free allowance. If access requires a purchase, the route is no longer a zero-payment option for that account.

Review the result before using another free attempt

Open the generated video and watch the event from beginning to end. A good opening frame can conceal an object that deforms or changes direction a few seconds later.

Ask whether the result completes the brief. Does the boat remain a folded paper object? Does it travel in the requested direction? Does the camera stay still? Is there unwanted lettering? If sound is important, listen to it; an audio track can exist without containing the dialogue or effects you need.

Change the instruction that addresses the actual failure. If the camera moves, make that constraint clearer. If the subject looks wrong, simplify the scene or consider an image reference, recognizing that this changes the input route. Repeating the entire request with more adjectives is less informative than correcting one visible problem.

Once you accept a clip, use its available download option and open the saved file. Check the actual dimensions, duration, visible watermark and sound. Google's watermark rules vary: its help describes invisible SynthID and visible-watermark controls, with automatic visible marking for certain markets and tiers. Do not assume every free Flow download is logo-free.

What a verified free export can—and cannot—show

A separate October 21 example used a Vidu Free account to animate an existing courier-robot image. It completed a 5.042-second reference-animation export using an account trial, without a credit-balance reduction. That was an image-to-video operation, not a direct text-only scene generation.

The source clip was placed once between short title cards in Canva, using free assembly features and retaining the Vidu watermark.

A short finished example: a Vidu trial animation sits between title cards on a Canva Free timeline, exported at 1080 × 1350.

The useful lesson is to check the final file as well as the generation screen. A free model, an account trial and a free editor can contribute different parts of a result. Name the contribution correctly rather than calling the whole process one free text-to-video generation.

If the free route stops, identify the reason

Free queues can be busy. If a request cannot start, save the prompt and references, then continue with an available model or the still-image assembly route rather than rebuilding the project from scratch.

For broader choices, use the free AI video project comparison. If visible markings are the decisive issue, the free video and watermark guide treats that as a separate export decision.

Start with one scene and a short prompt, then build on the result you like. Use AI Video with Seedance 2.0 to explore original motion with free daily credits; choose a presenter workflow when the main job is delivering a spoken script.