I tried out Kling 3.0 on Pollo AI! Testing the workflow from image generation to video creation

Introduction

Hello, I'm Nekoko🐾

This time, I'll be summarizing my experience creating examples using Pollo AI and focusing on one of the latest models available, Kling 3.0.

Recently, the number of AI video models has been increasing,
making it easy to wonder, "Which one should I actually use?"💦

In this article, I tried not only creating videos with Kling 3.0, but also generating the reference images used for the videos within Pollo AI.

Being able to go from image creation to video generation in the same place
makes the workflow much easier😊

In this article, I will look at the following based on actual examples after creating short videos:

• What kind of atmosphere is easy to produce

• What kind of prompts are easy to get results from

• Whether the flow from reference image to video is easy to use

I will examine these points based on the examples.

👉 Pollo AI Official Website

Kling 3.0 points I was curious about

First, when looking at the Pollo AI introduction page for Kling 3.0, these are the points that personally caught my attention.

• Multi-shot expressions

• Subject consistency

• Native audio (sound)

• Text retention (signs, logos, etc.)

• Video generation up to 15 seconds

• Storyboard-like control

When I actually tried it out, I got the impression that Kling 3.0 really captures the intent of the prompt well.
The official version also enhances features that make it easier to reflect detailed instructions, such as narration control, storyboard control, and text retention.

Theme of this example

The theme of this example is a video that emphasizes atmosphere.

Here are the three points we will look at this time👇

• Natural movement (does it break down suddenly?)

• Cohesion of the overall video (color, light, and atmosphere)

• Ease of use when converting from a reference image to a video

First, create a reference image

Rather than starting directly with video, if you solidify the atmosphere with a reference image first, it is easier to decide the direction when you turn it into a video.

This time, I completed this entire flow just within Pollo AI.

First, I will create the reference image to be used for the first video.
Basically, any image generation model can be used to create the reference image.
Any model you like is fine😊

Image of reference image ①

• Light reflection, reflections

• Atmosphere

Generation prompt for reference image ①👇

雨上がりの日本の都会の街（東京の繁華街のような雰囲気、昼〜夕方）。
主役として高校生くらいの女子生徒がはっきり見える（中景、全身または半身）。
透明な傘をさして歩いている。制服姿（自然な日本の高校生の制服）で、人物はシルエットではなく、顔や服が柔らかい光で見える。
濡れた道路に街の光や看板の光が反射している。広めの通り、ビルのガラス、都会的な店の外観、信号や街灯など、現代的な日本の都市らしい要素を入れる。
シネマティック、自然な色、少し明るめの露出、落ち着いた雰囲気、すっきりした構図。動画化しやすいシンプル構成。読める文字は入れない。

This time, I wrote the prompt in Japanese.

※AI generation does not guarantee the exact same result every time, even if you use the same prompt. Please view this example as just one possibility.

👇Generated reference image

At this point, being able to see "what kind of atmosphere I want the video to have" made the next video generation much easier to proceed with.
I could have just turned it into a video normally, but I wanted to experiment a little😊
I will turn this generated image into a line drawing to make it look like anime👇

Prompt used for line drawing conversion👇

アップロードした絵を線画にしてください。
アニメ風の線画にしてください。

I will create a video from this image using Kling 3.0😊

I turned that image into a video with Kling 3.0

Next, I will use the reference image I just created to make a video about 15 seconds long with Kling 3.0.

This time, I prioritized conveying the atmosphere over flashy movements.

Concept for Example 1

• Slow camera movement

• Reflections and glare

• Calm atmosphere and mood

• A slightly unusual video

Generation prompt for Example 1👇

女の子が傘をさして歩いている。
徐々に着色されていき、最後には綺麗なアニメになる。

It is a very short Japanese prompt.
Even with a prompt like this, Kling 3.0 creates a solid video.
This is why I felt Kling 3.0 is quite smart.

👇Generated video

My first impression was that the human movement felt quite natural.
Also, even with a somewhat vague prompt about adding color little by little while moving, it seems to have captured my intent well.
The reflection in the puddle in the background is also nice.
It even includes sound that wasn't written in the prompt.
If I had to ask for more, I would have liked it to add a bit more color😊
But it's amazing that it can create a full-fledged animation from a line drawing!

I tried making another pattern

Next, I tried the workflow of turning a reference image into a video one more time with a slightly different atmosphere.

Just like the first one, I created the reference image first and then turned it into a video with Kling 3.0.

Concept for Reference Image 2

• Brighter atmosphere

• A slightly more rhythmic presentation

• A different vibe from the first one

Prompt for generating reference image ②👇

混み合う日本の通勤電車の車内。
立っている乗客が多く、つり革を持っている人が並んでいる。奥行きのある車両構図。
主役として高校生くらいの女子生徒が中景ではっきり見える（半身〜全身）。
自然な日本の制服姿。人物はシルエットではなく、顔や髪型が分かる。
車内の手すり、つり革、座席、ドア、窓など日本の電車らしい要素を入れる。
アニメ風のフルカラーイラスト、やわらかい光、自然な配色、背景は少し整理して主役が埋もれないようにする。
雰囲気重視、すっきりした構図、動画化しやすいシンプル構成。

👇Generated reference image

Video generation prompt👇

同じ女子生徒（参照画像の人物）を主人公にしたアニメ風マルチショット動画。
顔・髪型・制服（紺ブレザー、白シャツ、赤リボン）・黒いリュックを全ショットで維持する。自然な日本の朝の通勤電車の雰囲気。カメラは安定、自然な動き。各ショットは短く、1ショットにつき1つの動きだけを見せる。

ショット1：混み合う日本の通勤電車の車内。主人公の女子生徒が中景ではっきり見える。つり革、乗客、車内の自然な小さな揺れ。主人公は静かに立っている。
ショット2：電車の窓から見える都会の流れる風景。ビルや道路が自然に流れる。見やすく安定した映像。反射は少しだけ。
ショット3：日本の駅ホーム。電車はすでにホームに停車している（動いていない）。主人公の女子生徒が同じ制服・同じリュックのまま、ホームを自然に歩いていく。背景に停車中の電車と開いたドアが見える。朝の通勤の空気感。

アニメ風フルカラー、自然な配色、やわらかい光、物語性重視。

👇Generated video

As expected, the human movement is quite natural!
It also handles multi-shots very well!
Even without writing detailed instructions about sound, it included ambient noise😊

Since the prompt was more complex than the first one, there are some points that bother me if I look closely, but still, being able to do this much with Kling 3.0 is amazing👏

Ease of use of Kling 3.0 that I felt after using it

I will summarize my impressions from trying it out here.

I felt it was quite easy to create even a 15-second video, which is a bit long for an AI-generated video.
It seems to understand even slightly abstract prompts in Japanese that convey a mood, making it easy to get good results😊

Pollo AI benefits of using it

After focusing on Kling 3.0 this time, what I felt again is that video generation can be completed relatively easily just on Pollo AI.

Although I am focusing on Kling 3.0 this time, with Pollo AI you can

• Create a reference image

• Turn it directly into a video

• Use other features if necessary

It was helpful to be able to proceed with that flow in the same place.

Since you don't have to switch between tools, it's a big plus that the tempo of prototyping doesn't drop😊

It's not about "which is the strongest model?", but rather that having a place where it's easy to try things out according to what you want to create is what makes
Pollo AI easy to use.

Summary

This time, I tried out Pollo AI's Kling 3.0 by creating examples, from making reference images to generating videos.

Personally, I felt it was easy to use as an entry point for video, as it's easy to create both short and slightly longer videos🐾

Moving forward, I would like to gradually test out

• audio features (native audio)

• text retention

, and other areas that I didn't get to cover much this time.

Bonus

Video effect:Beast Chase👇

You can choose from a bear, dinosaur, or octopus for the animal chasing you😊
It makes for a fun video😆

Links & Related Articles

👉 Pollo AI Official Website

👉 Pollo AI "Kling 3.0" Introduction Page
https://pollo.ai/ja/m/kling-ai/kling-3-0

🐾 Check out these related articles too!

Here is the link to the note magazine containing articles where I've introduced Pollo AI so far👇

Thank you for reading this far🐾

Hashtag suggestions

#PolloAI #Kling3_0 #AIVideoGeneration #ShortVideo #GenerativeAI
#VideoProduction #AITools #ContentCreation #VideoMaking
#AnimeStyle #MultiShot #Workflow