EN

You're Not Behind (Yet): How to Start Making AI Videos in 14 Minutes

#AI Tools
2026-09-30 11:4316 min read

Content Introduction
Ask Questions

The video explains a faster workflow for making AI videos using Higsfield. It presents three methods: text-only prompts, image-to-video with reference frames, and full control using both start and end frames. The creator demonstrates multiple examples, including pottery, cooking, a forge scene, a personal character sheet, running in a yellow room, and a swimming race. Across these examples, the key lesson is that better results come from precise prompts, timing cues, sound design, and locking in visual details like composition, lighting, and character identity. The video argues that building from an image first gives more control than describing a scene with words alone, and it promotes Higsfield and its new model with a discount offer.

Key Information

  • The speaker argues that people making AI videos the traditional way are falling behind, while a faster, more controlled workflow can produce better results.
  • The workflow is centered on using one platform, Higsfield, instead of switching between multiple tools.
  • A prompt-building bonus tool can convert a simple description into a model-specific, ready-to-use prompt, which improves output quality.
  • Choosing the correct model, aspect ratio, duration, and resolution matters; the speaker repeatedly uses Seedance 2.5 and Nano Banana Pro with settings like 16:9, 1080p, and 2K.
  • Strong AI video results come from writing shots in timed segments, even for single-shot scenes, so the model has a clear beginning, middle, and end.
  • For more control, the speaker recommends creating a starting image first, then using it as the reference frame for video generation.
  • When using reference images, detailed visual descriptions such as lighting, composition, texture, and realism cues help the model preserve the intended look.
  • To keep a character consistent across scenes, the speaker uses a character sheet or reference sheet and carries those details into both image and video prompts.
  • The speaker demonstrates several examples: a potter shaping a vase, a chef plating food, an astronaut scene, a blacksmith forge, a woman on a treadmill, and a swimming race.
  • The most controlled method is to define both a start frame and an end frame, then let the model generate the motion in between.
  • Sound, timing, and motion cues are emphasized as more important than visuals alone for making scenes feel complete and engaging.
  • The video ends with a promotion for Higsfield and a discount on the Seedance 2.5 model.

Timeline Analysis

Content Keywords

Higsfield

An all-in-one AI video platform used throughout the workflow to generate videos, images, and character-based scenes without switching tools. The speaker emphasizes that Higsfield combines multiple creative tools in one place, making the process faster and more controlled.

Seedance 2.5

The video generation model selected in the workflow for creating motion from prompts and reference frames. It is used for multiple scenes, including pottery, cooking, sports, and underwater action, with different durations and resolutions.

Nano Banana Pro

The image generation model used to create precise start and end frames before animating them in video. The speaker uses it to generate realistic reference images, character sheets, and controlled compositions for smoother AI video results.

AI video workflow

A structured process for making AI videos using prompts, timestamps, reference images, and frame locking. The speaker explains three methods: text-only prompt generation, image-to-video, and start/end frame control for the most precision.

Prompt engineering

A key technique in the script where the speaker stresses writing detailed prompts with timing, motion cues, lighting, composition, and specific visual details. Better prompts produce more realistic movement, stronger continuity, and fewer generic results.

Image-to-video generation

A method where a generated or reference image is used as the starting point for video creation. The speaker shows that a strong first frame helps control identity, composition, and realism, while the video prompt adds motion.

Character consistency

Maintaining the same face, body, and identifying details across multiple generated scenes. The speaker uses reference sheets and detailed descriptions like facial asymmetry and skin texture to keep the character recognizable.

Start frame and end frame control

A more advanced workflow where both the first and last frames are locked before generating the video. This gives the model clear boundaries and allows it to invent the motion in between with much more control.

Realistic AI visuals

The script highlights realism through details like clay texture, sparks, water physics, sweat, shadows, reflections, and natural body movement. The speaker repeatedly shows how specifying visual details improves the final output.

AI video monetization

The closing call-to-action promotes Higsfield as a way to make AI videos more efficiently, with a discount on the newest model and a link in the description. The overall message is to create better AI videos faster using a streamlined workflow.

Related questions&answers

What is the main workflow described for making AI videos?

The workflow is to generate a strong image first, then turn that image into a video, and for the most control, create both a start frame and an end frame before generating the motion in between.

Why does the speaker say the normal way of making AI videos is falling behind?

Because people using outdated workflows are slower and get less consistent results, while a more structured image-to-video workflow can produce videos faster and more predictably.

What platform does the speaker use for the whole process?

The speaker uses Higsfield for both image and video generation instead of switching between multiple tools.

What is the Higsfield bonus package?

It is a tool the speaker made that turns a simple description into a full ready-to-use prompt, tailored to the model you choose.

Why does the speaker say you should specify the model before generating the prompt?

Because the prompt will be tuned for that exact model rather than being a generic version, which helps the output better match what you want.

What model does the speaker use for video generation?

The speaker uses Cedance 2.5 for video generation.

What image model does the speaker use for making reference images?

The speaker uses Nano Banana Pro for image generation.

Why does the speaker recommend writing prompts in timed parts?

Because timed prompts help the model understand how the scene should develop across the clip, giving the video a beginning, middle, and end even in a single shot.

How does the speaker make a single-shot pottery video feel structured?

By describing the action in timed stages: the wheel starts spinning, the clay rises and narrows in the middle, and the vase finishes at the end.

What is the advantage of using an image as the starting frame?

It locks in the visual style and composition at the beginning so the video can move from a clear reference instead of guessing the look from text alone.

Can a starting image control the entire video?

No, the starting image only fixes where the video begins, not where it can go. The motion still depends on the video prompt.

How does the speaker create a realistic blacksmith scene?

They generate a detailed start image of a blacksmith at a forge, then use that image in the video tab with a prompt that guides the motion and preserves the lighting and composition.

Why does the speaker create a character sheet from a photo?

To keep the same person consistent across multiple videos by giving the model a clear reference for the face and body.

What details help the AI keep the character looking realistic?

The speaker includes specific details like visible pores, slight asymmetry, and even a small pimple, which help preserve natural-looking identity.

Why is sound important in these AI videos?

The speaker says sound and timing matter more than the picture alone, and that a quiet scene with strong sound can be more effective than a visually busy one.

What is the benefit of locking both a start frame and an end frame?

It gives the model clear boundaries for the scene, reducing guesswork and making the final motion more controlled and predictable.

How does the speaker approach a close-up-to-wide-shot running scene?

They use a close-up start frame and a wide end frame, then describe motion cues like upper-body bounce, shoulder movement, hair motion, and heavy breathing.

Why does the speaker mention describing the camera and framing so specifically?

Because the model only animates what is actually described, so matching the motion cues to the shot size helps the video feel believable.

What kind of visual style does the speaker use for the swimmer scene?

They describe a muted teal and blue palette with soft daylight from high windows to give the model a clear cinematic look.

What is the speaker's final recommendation for getting better AI videos?

Use a repeatable workflow: create a strong image first, use reference frames when needed, write precise motion prompts, and keep image and video generation within one platform if possible.

More video recommendations

Share to: