blackdark
Content

5 faceless niches with AI: the style prompt, the scenes and the step-by-step

The 5 fastest-growing faceless video niches right now (oil-painted history, wolf fables, the stretchy couple, cube slicing and first-person “what if…?” POV), each with its style prompt, 3 scene prompts, the motion prompt and the free route with Google Flow.

By Updated on 12 min read
Jump straight to the 5 niches and their prompts

In the carousel you saw 5 faceless video niches that are growing right now, and nobody shows their face in any of them. Here’s what I promised: the style prompt for each one, 3 copy-paste scene prompts, the motion prompt and the step-by-step to get your first video out.

The example images in this guide were made with these exact prompts (Nano Banana Pro, 9:16 format).

Note

Who does what. The image for each scene comes from Nano Banana (in Gemini, Google Flow or kie.ai). The motion comes from an image-to-video model: Veo in Google Flow, or Kling. The voice comes from ElevenLabs or from Flow itself. The edit, in CapCut.

From idea to faceless video

  1. Pick a niche

    And don’t change it for your first 5 videos.

  2. Lock the style

    Paste the style prompt in front of every scene.

  3. Short script

    5-8 scenes, one line per scene, hook in the first second.

  4. Images in Nano Banana

    Style + scene, 9:16 and always NO TEXT.

  5. Motion in Veo or Kling

    5-8 seconds per image with the motion prompt.

  6. Voice, sound and edit

    Voiceover, sound effects and big captions in CapCut.

The 5 faceless niches at a glance

#NicheStyleReference accountFollowersMost-viewed reel
1Dramatized historyRealistic oil painting, slow motion@null.histories995K178K
2Wolf fablesCartoon noir, psychology quotes@quietwolfzz1.3M10.3M
3Stretchy coupleClaymation-style 3D, no-dialogue gags@mr._stretchy1.3M27M
4Cube slicingSatisfying 3D, “0 cuts vs 1,000”@infiniteslice0173K45.2M
5First-person “what if…?”Realistic POV, absurd science@realityrendered__72K1.5M

Instagram data checked on November 8, 2026; the “most-viewed reel” is the peak among each account’s ~30 latest posts. Accounts are cited only as format references.

Step by step (works for all 5 niches)

  1. Pick ONE niche and don’t change it for 5 videos. What grows these accounts is repetition: same character, same shot, same structure.
  2. Lock the style: copy the niche’s style prompt and always paste it in front of every scene. If there’s a recurring character (niches 2 and 3), generate their character sheet first and pass it as a reference image in every scene.
  3. Short script: 5-8 scenes, one line per scene. The strongest line goes first: that’s your hook.
  4. Images: generate each scene with the style prompt + the scene prompt, in 9:16 format. The style prompt already includes “NO TEXT” so the AI doesn’t add weird lettering.
  5. Motion: animate each image for 5-8 seconds with Veo (in Google Flow) or Kling, using the niche’s motion prompt.
  6. Voice and sound: voiceover (ElevenLabs or Flow’s voice) and sound effects. In niches 3 and 4, the ASMR sound is half the video.
  7. Edit in CapCut: a cut on every line, big captions and 15-40 seconds total. Post and repeat.

How much it costs: free route vs paid route

RouteImageVideoCost
FreeNano Banana in Google Flow or GeminiVeo 3.1 in Google Flow$0: 50 credits a day (Lite = 10 per video, Fast = 20)
Cheap paidnano-banana-pro on kie.aiKling 3.0 std on kie.ai≈ $0.07 per second of video (a 5 s clip ≈ $0.35)

With Google Flow’s 50 daily credits you get about 5 clips with Veo 3.1 Lite or 2 with Veo 3.1 Fast a day: enough to get your first test video out in a day or two. Credits don’t roll over to the next day, and at peak times Flow may not let you generate video without a subscription (images still work). Source: Google Flow’s credits help page.

Tip

The Claude shortcut. Paste your niche’s style prompt into Claude and ask: “Write me a 6-scene script about [YOUR TOPIC] and, for each scene, the scene prompt in English in the same format as these examples.” You get the whole video back, ready to generate.

1. Dramatized history (realistic oil painting)

Three AI-generated example scenes in realistic oil painting style: a Roman general before a burning senate, an Egyptian queen in her throne hall and a sailor at the helm in a storm
The 3 examples for niche 1, generated with the prompts below

How it works: a real moment in history told in the first person by the person who lived it. 6-8 scenes of 5 seconds, one line per scene, a deep slow voice, and an ending with a twist (“and nobody saw it coming”).

Style prompt · Dramatized history
Vertical 9:16 frame, a single still from a short-form AI video. NO TEXT, no letters, no captions, no subtitles, no logos, no watermarks, no UI, no borders.
STYLE: dramatic realistic oil painting, visible thick brushstrokes, Rembrandt chiaroscuro lighting, cinematic slow-motion feel, rich palette of deep umber #3b2414, burnt gold #c8902e, crimson #8e1b1b and smoky black #0d0b0a, volumetric god rays and drifting dust, museum-quality historical painting, original characters (not real portraits).
Scene 1 · Roman general
a Roman general in a crimson cloak and bronze armor stands alone on the steps of a burning senate building at dusk, shouting a defiant order, embers floating in the air, crowd of soldiers blurred in the background, low camera angle.
Scene 2 · Egyptian queen
an Egyptian queen with a gold headdress turns her head toward the viewer inside a torch-lit throne hall, expression of cold fury, papyrus scrolls falling from a table, golden light from braziers, close-up medium shot.
Scene 3 · Sailor in the storm
a 15th-century sailor with a grey beard grips the wheel of a wooden caravel during a violent storm at night, a blood-red moon breaking through the clouds, huge waves, lantern light on his face.
Motion · image to video (5 s)
Slow cinematic push-in, dust and embers drifting, the character turns slowly toward the camera and speaks with a deep voice, oil painting texture preserved, 5 seconds.

2. Wolf fables (cartoon noir)

Three AI-generated example scenes in cartoon noir style: a wolf in a suit at a bar counter, under a street lamp in the rain and in an office with a chess board
The 3 examples for niche 2, generated with the prompts below

How it works: first generate the wolf’s character sheet (front, profile, 3/4 and 4 expressions) and use it as a reference in every scene. Each video is a psychology lesson (for example, “the one who talks the most knows the least”) in 3-5 scenes.

Style prompt · Wolf fables
Vertical 9:16 frame, a single still from a short-form AI video. NO TEXT, no letters, no captions, no subtitles, no logos, no watermarks, no UI, no borders.
STYLE: cartoon noir illustration, bold clean ink outlines, cel shading, moody 1940s detective comic look, original anthropomorphic grey wolf character with amber eyes wearing a charcoal pinstripe suit and fedora, palette of midnight blue #101a2b, smoky teal #2f4f5a, warm lamplight amber #e3a03a and black #050505, rain and neon reflections, cinematic composition.
Scene 1 · At the bar
the wolf sits calmly at a bar counter at night swirling a glass of water, a nervous cartoon fox in a cheap suit next to him sweating, the wolf giving a knowing half-smile, warm lamp above. Behind them only a plain dark wooden wall and bottles without labels: no windows with signs, no neon lettering, no writing anywhere.
Scene 2 · Under the street lamp
the wolf stands alone under a street lamp in the rain at night, collar up, looking over his shoulder at the viewer with a calm confident expression, puddles reflecting the lamp light. Plain brick walls and blank unlit windows only: no shop signs, no neon lettering, no writing anywhere.
Scene 3 · In the office
the wolf leans back in a leather chair in a dark office, hands steepled, a chess board with one piece knocked over on the desk, venetian blind shadows across the room.
Motion · image to video (5 s)
Subtle noir animation, rain falling, lamp flicker, the wolf slowly raises his eyes and smirks, smoke drifting, comic-book shading preserved, 5 seconds.

3. Stretchy couple (claymation-style 3D)

Three AI-generated example scenes in claymation-style 3D: a man with stretchy arms grabs a jar in the kitchen, hangs laundry from the sofa and steals a slice of cake
The 3 examples for niche 3, generated with the prompts below

How it works: lock the 2 characters with a character sheet each. One gag per video (8-10 seconds) and no dialogue: sound rules (rubbery stretches, plates, laughs). The gag has to land even with the sound off.

Style prompt · Stretchy couple
Vertical 9:16 frame, a single still from a short-form AI video. NO TEXT, no letters, no captions, no subtitles, no logos, no watermarks, no UI, no borders.
STYLE: 3D stop-motion clay animation look, plasticine characters with fingerprints and soft subsurface skin, Aardman / Pixar-like charm, two original characters: a lanky man with extremely long rubbery stretchy arms in a mustard sweater and a cheerful short woman with curly red hair and a green apron, cozy domestic set, warm soft key light, palette of cream #f3e6cf, mustard #d9a32b, sage green #8fae7b and terracotta #c4623a, shallow depth of field.
Scene 1 · The kitchen jar
in a small kitchen the man stretches his arm across the whole room to grab a jar from the top shelf while the woman cooks spaghetti and laughs, comedic gag moment, eye-level shot.
Scene 2 · Laundry from the sofa
in a living room the man uses his super long stretched arms to hang laundry on a line from the sofa without getting up, the woman with hands on hips looking at him annoyed but amused.
Scene 3 · The slice of cake
at a dining table the man's stretchy arm sneaks across the table to steal a slice of cake from the woman's plate while she is looking away, close comedic framing.
Motion · image to video (8 s)
Stop-motion clay animation at 12 fps feel, the man's arm stretches like rubber across the room in one smooth comedic motion, the woman reacts with a surprised laugh, 8 seconds, no dialogue.

4. Cube slicing (satisfying 3D)

Three AI-generated example scenes in satisfying 3D: a pineapple, a layered jelly cake and an apple sliced into perfect cubes on a wooden board
The 3 examples for niche 4, generated with the prompts below

How it works: the same shot and the same board in every video (it’s the channel’s signature). Structure: 0 cuts → 10 → 100 → 1,000, with a big on-screen counter you add yourself in the edit. An ASMR cutting sound on every hit.

Style prompt · Cube slicing
Vertical 9:16 frame, a single still from a short-form AI video. NO TEXT, no letters, no captions, no subtitles, no logos, no watermarks, no UI, no borders.
STYLE: hyper-satisfying 3D render, clean studio lighting, glossy physically based materials, objects sliced into perfect tiny cubes, oddly satisfying ASMR aesthetic, plain wooden cutting board on a dark charcoal #1a1a1a background, crisp macro detail, juicy cross-sections, top-down three-quarter angle, same camera framing in every shot.
Scene 1 · Pineapple
a large chef knife slicing a whole pineapple, the pineapple splitting into hundreds of perfect yellow #f2c230 juicy cubes that cascade on the board.
Scene 2 · Jelly cake
a knife cutting a block of colorful layered jelly cake into tiny perfect cubes, translucent layers of red #e23b3b, orange, green and blue glowing in the light, cubes tumbling.
Scene 3 · Apple
a knife slicing a big glossy red apple into a perfect grid of tiny cubes, white flesh visible, the apple holding its shape for a moment before collapsing into cubes.
Motion · image to video (6 s)
Satisfying slow-motion cut, the knife glides down and the object separates into perfect tiny cubes that tumble softly on the board, crisp ASMR feel, locked camera, 6 seconds.

5. First-person “what if…?” (realistic POV)

Three AI-generated example scenes in realistic POV: a kayak facing a megatsunami in a fjord, a skyscraper rooftop facing a sandstorm and an astronaut with a giant Moon
The 3 examples for niche 5, generated with the prompts below

How it works: open with the question on screen (“What if you were in a kayak when a megatsunami hit?”). 4-6 POV scenes of 6-8 seconds that keep escalating; the last one is the wildest. A calm voiceover, in contrast with the visuals.

Style prompt · “What if…?” POV
Vertical 9:16 frame, a single still from a short-form AI video. NO TEXT, no letters, no captions, no subtitles, no logos, no watermarks, no UI, no borders.
STYLE: photorealistic first-person POV shot, as if filmed with a wide action camera, the viewer's own hands visible at the bottom of the frame, epic scale, cinematic realism, natural lighting, slight motion blur, Unreal Engine 5 realism, palette of steel blue #3b5a78, glacier white #e8eef2 and storm grey #5a5f66.
Scene 1 · Kayak and megatsunami
POV in a small red kayak paddling in a narrow fjord, the viewer's hands holding the paddle, a colossal tsunami wave rising between the mountains ahead, spray in the air.
Scene 2 · Rooftop and sandstorm
POV standing on top of a skyscraper roof edge at sunset, the viewer's sneakers visible, the city below being swallowed by a gigantic sandstorm wall rolling in, orange dust #c9743a.
Scene 3 · Astronaut and the Moon
POV floating in a spacesuit outside a space station, the viewer's gloved hands reaching toward a tether, Earth below and the Moon appearing enormous, unrealistically close, filling half the sky.
Motion · image to video (6-8 s)
First-person POV, handheld camera shake, the viewer's hands react as the giant event approaches, wind and spray hitting the lens, realistic physics, 6-8 seconds.

Mistakes that sink a faceless niche

  • Changing style every video. The algorithm and your audience need to recognize you in one second. Same style prompt, always.
  • Letting the AI write in the image. You get warped text. Ask for “NO TEXT” and add captions in the edit.
  • Clips that are too long. Past 8 seconds the character starts to deform. Better to use more short scenes.
  • A different character in every scene. In niches 2 and 3, the character sheet as a reference is a must.
  • Copying the reference account. Use it to understand the format, not to clone its videos. Your topic, your character and your voice.

Note

Heads-up. YouTube and Meta penalize repetitive mass-produced content. We don’t promise virality or monetization: these are a base for you to make something of your own.

Tip

Want it without editing anything by hand? My Fábrica de Reels pack builds the whole reel with Claude Code: script, voice, animation and captions. It’s optional: everything in this guide works without it.

FAQ

Yes, to get started. Google Flow gives everyone 50 credits a day, with or without a subscription, and they don’t roll over. A video with Veo 3.1 Lite costs 10 credits and one with Veo 3.1 Fast costs 20, so you get about 5 Lite clips or 2 Fast clips a day. At peak times Google may not let you generate video without a subscription; images still work.

On kie.ai, Kling 3.0 std costs about 14 credits per second (≈ $0.07/s), so a 5-second clip comes out at around $0.35. Images with nano-banana-pro are extra and cost cents.

The one you can keep up for 30 days. If you like storytelling, 1 or 2. If you don’t want to write scripts, 3 or 4, which live on the gag and the sound. If you’re into science and epic stuff, 5. What matters is not switching niches during your first videos.

Image and video models follow style details better in English. The voiceover and captions can be in whatever language you post in.

We don’t promise that. YouTube and Meta penalize repetitive mass-produced content. Use these as a base to make something of your own: your character, your stories, your edit.

Keep digging into the same topic.

Share
Newsletter

Get the next guides in your inbox

AI and marketing ideas and resources, no filler. What works and how to apply it.

Ideas and resources, no spam. Unsubscribe anytime.

Finding this guide useful?

Subscribe