How to Make Those AI Fruit Videos (Drama, Talking, ASMR)

Those AI fruit videos (talking fruit soap operas, fruit characters with faces, glass fruit being sliced for ASMR) are made with text-to-video or image-to-video models that generate sound along with the picture, such as Veo 3.1 and Seedance 2. The usual workflow is: design the fruit character as a still image, animate it in short 5-8 second clips with the dialogue written into the prompt, then join the clips. On VdoBloom a 32-second, four-clip fruit drama costs about 170 credits on Veo 3.1 Fast, or about 115 credits on Seedance 2 Mini.

Disclosure: VdoBloom is our platform.

What the three kinds of AI fruit video actually are

People searching “how do people make those AI fruit videos” usually mean one of three formats. Each one needs a slightly different setup, so it helps to know which you are making before you spend any credits.

• Fruit drama. Fruits with faces, arms and outfits act out a soap-opera plot: a strawberry catches her banana boyfriend with a peach, someone gets slapped, someone cries at a kitchen table. These are multi-clip stories with dialogue, and the hard part is keeping each fruit looking the same in every clip.
• Talking fruit. One fruit character speaking to camera: a lemon giving life advice, an avocado reviewing breakfast. This is a single clip or a few clips with one character and a voice.
• Fruit cutting ASMR. A knife slicing fruit that is made of glass, crystal, jelly or lava, with close-up crunch and clink sounds. No characters, no story, just one satisfying 8-10 second shot where the sound matters as much as the picture.

What all three have in common is audio. A silent fruit drama does not work, and ASMR without the slicing sound is pointless. That is why these videos took off once video models started generating sound together with the picture instead of leaving you to add it afterwards.

Which AI models people use for fruit videos

You do not need one special “fruit video app”. Any strong video model with native audio can do it. On VdoBloom the practical choices are:

Model	Good for	Sound	Example cost on VdoBloom
Veo 3.1 Fast	Fruit drama dialogue, talking fruit	Generated with the clip	40 credits per 8-second clip
Seedance 2 Mini	Cheap drafts, longer clips up to 15 s	Generated by default	17 credits for 5 s at 720p, 26 for 8 s
Seedance 2 Fast	Sharper ASMR close-ups	Generated by default	50 credits for 5 s at 720p, 100 for 10 s
Seedance 2	Best detail up to 4K	Generated by default	82 credits for 5 s at 720p

Prices are what VdoBloom charges as of September 2026; the exact number is always shown on the Generate button before you click it. For the fruit character stills you will also use an image model. GPT Image 2 costs 3 credits at 1K and 5 credits at 2K, and Nano Banana Pro costs 18 credits, and both are good at following a detailed character description.

Two rules of thumb: use Veo 3.1 Fast when the clip has dialogue and you want the voices to sound natural, and use Seedance 2 Mini for drafts and for scenes without speech, then regenerate only the best ones on a stronger model.

Step by step: make an AI fruit drama on VdoBloom

This is the workflow for a four-scene fruit drama of about 32 seconds, the length most of these clips run on TikTok and Reels.

1. Write the plot as four beats. Keep it simple: setup, reveal, confrontation, twist. One line of dialogue per beat is plenty, since each clip is only 5-8 seconds and a character can say about one sentence in that time.
2. Design each fruit character as a still image. Go to Images → Generate, pick GPT Image 2 or Nano Banana Pro, and set the ratio to 9:16. Describe the character in detail and save the result. You will reuse this exact description in every video prompt.
3. Make one starting frame per scene. Open Images → Edit, upload your character image(s) as references, and ask for the scene: “the same strawberry character standing in a kitchen doorway, shocked, 9:16”. This locks the look before any video credits are spent.
4. Animate each frame. Open Video → Image to Video, upload the scene frame, pick Seedance 2 Mini or Seedance 2 Fast, set 9:16 and 8 seconds, and write the action and dialogue in the prompt. When two fruit characters share a scene, Seedance 2 models on this tab can instead take up to 9 reference images (for example one per character) that you tag as @Image1, @Image2 in the prompt; this reference mode replaces the single starting frame rather than adding to it.
5. Or skip the stills and go straight to text. On Text to Video with Veo 3.1 Fast, you can write the whole scene, character description and dialogue in one prompt. It is faster but the characters drift more between clips.
6. Join the clips. Download the four clips and put them in order in any editor (CapCut, InShot, Premiere). Add captions there, because burned-in AI text is often misspelled.

Sample prompts you can copy

Character still (image model):

“A cute anthropomorphic strawberry woman, glossy red skin with golden seeds, big expressive brown eyes with long lashes, small green leaf crown like a hairstyle, thin cartoon arms and legs, wearing a tiny pink apron. Pixar-like 3D render, soft kitchen lighting, full body, plain background, 9:16.”

Scene clip (video model):

“The strawberry woman from the image stands in a sunny kitchen doorway. She sees a banana man holding hands with a peach woman at the table. She gasps, drops a teacup that shatters on the floor, and says angrily: ‘Kevin, after everything we’ve been through?’ Dramatic soap-opera music sting, slow push-in on her face. Keep her appearance exactly as in the image.”

Talking fruit (single clip):

“Close-up of a grumpy lemon character with a wrinkled yellow face and tiny sunglasses, sitting on a kitchen counter, talking directly to camera in a deep sarcastic voice: ‘You want me in your water? Earn it.’ Shallow depth of field, warm morning light, 9:16.”

Put dialogue in quotes and name who says it. Models are much more likely to produce the right line in the right mouth when the prompt is explicit. If a character says the wrong words, shorten the line; long sentences get garbled first.

Step by step: glass fruit cutting ASMR

ASMR fruit cutting is the easiest of the three because it is one shot with no characters.

1. Open Text to Video.
2. Pick Veo 3.1 Fast (40 credits per clip) or Seedance 2 Fast at 720p for 10 seconds (100 credits). For cheap test runs, Seedance 2 Mini at 720p is 33 credits for 10 seconds.
3. Set 9:16 for Shorts, TikTok or Reels.
4. Write the prompt with the material, the knife action, the camera and the sound all spelled out.

Sample ASMR prompt:

“Extreme macro close-up, a sharp chef’s knife slowly slices through a transparent glass kiwi on a dark wooden cutting board. The glass kiwi has green translucent flesh with tiny black glass seeds. Each slice separates cleanly and tips over. Crisp crunchy glass-cracking ASMR sound, soft tapping of the knife on the board, no music, no voice. Studio lighting, black background, slow motion, 9:16.”

Swap the material to get the other popular versions: “made of honey-colored jelly”, “made of crystal”, “made of molten lava rock”. Say “no music, no voice” explicitly, otherwise some models add a soundtrack that drowns out the crunch.

What a finished fruit video costs

These are real VdoBloom prices as of September 2026. Plan for some retries: a clip where the fruit grows a second face or the dialogue lands on the wrong character is normal, so budget roughly one extra generation for every two or three clips.

Project	What you generate	Credits
Fruit drama, 32 s, Veo 3.1 Fast	2 character stills (GPT Image 2 at 2K, 5 each) + 4 clips at 40	170
Fruit drama, 32 s, budget	2 character stills at 5 + 4 Seedance 2 Mini clips, 8 s at 720p (26 each)	114
Talking fruit, one 8 s clip	1 Veo 3.1 Fast clip	40
Glass fruit ASMR, 10 s	1 Seedance 2 Fast clip at 720p	100

New accounts get 10 free credits, which is enough to test a character still or two but not a full clip with sound. If you want to understand how credits map to money on each plan, read AI video credits explained.

Keeping the same fruit character in every clip

The fruit drama format falls apart when the strawberry has a leaf crown in scene 1 and a bow in scene 3. Three habits fix most of it:

• Start every clip from an image, not just text. Image-to-video keeps far more of the design than text-to-video, because the model is animating your picture rather than reinventing it.
• Reuse the exact same description. Copy the character paragraph word for word into every prompt. Changing “pink apron” to “apron” is enough to get a different apron.
• Give each character one strong visual hook. Sunglasses, a scar, a hat. Small details drift; big distinctive ones survive.

Our longer guide on AI video character consistency covers reference images and first-frame tricks in more detail.

Posting fruit videos: labels and originality

TikTok, YouTube and Instagram all ask creators to label realistic AI-generated content, and cartoon fruit is rarely mistaken for real footage, but check each platform’s current rules before you post. Fruit drama channels also get copied heavily, so the ones that last write their own storylines and keep a recurring cast rather than remaking whatever plot went viral last week. Content on VdoBloom goes through a content filter; fruit slapstick is fine, and anything sexual involving characters that look like children is blocked.

Frequently asked questions

What app do people use to make those AI fruit videos?

Most are made with a video model that generates sound, such as Veo 3.1 or Seedance 2, and then edited together in an ordinary editor like CapCut. VdoBloom gives you both model families in one place, plus image models for designing the characters first.

How to make those AI fruit videos for free?

Fully free is hard, because video with sound is expensive to generate. VdoBloom’s 10 free signup credits cover a couple of character images; a Seedance 2 Mini clip at 480p for 5 seconds is 8 credits, which also fits, but for a full multi-clip drama you will need paid credits.

What is a good AI fruit video prompt?

Describe the character (fruit, face, outfit, style), the setting, one action, one line of dialogue in quotes with who says it, the camera move, and the sound. Keep each clip to one beat of the story.

How do I make AI fruit cutting ASMR videos?

Use text-to-video with a macro close-up prompt: the fruit material (glass, crystal, jelly), the knife action, a dark background, and the exact sounds you want, plus “no music, no voice”. One 8-10 second clip is the whole video.

How do they keep the fruit characters looking the same?

They design each character as an image first, then animate that image in every scene with the same written description, instead of generating each clip from text alone.