# Talking animal sitcom with AI: Barbary macaques speaking Darija

> The format: animals in clothes act out a short domestic joke and speak real lines, lip-synced, in a dialect. We made these clips in September 2026 with invented Barbary macaques, a species of the Atlas mountains, and dialogue written in Darija, the language we chose for the sketch. First a single macaque in a kitchen: a doctor tells him to cut down on bananas, so he cuts one in two and eats both halves. We drew that joke as a 6-panel board, then ran the same board on Gemini Omni Flash 1.1 and on Seedance 2.5. Then a two-macaque sofa scene, where we ran ten takes and kept one. Each example below has the exact prompt as sent, an English gloss of the dialogue, the settings, and the cost of the same run in Twistavo today.

Canonical page: https://twistavo.com/prompts/talking-animal-sitcom-ai-video
Updated: 2026-11-18

## Setup

- Model: Gemini Omni Flash 1.1, Seedance 2.5 and Seedream 5.0 Pro
- Settings: Every clip is vertical 9:16. The kitchen board is one Seedream 5.0 Pro image from text only. The board was then cut into separate panel images, and those panels went in as reference images: five for Gemini Omni Flash 1.1, six for Seedance 2.5. Gemini wrote the voice itself from the dialogue in the prompt (native audio). For Seedance we gave a separate voice file as the audio reference and asked it to lip-sync to it. The sofa scene uses one image of the two characters as its only reference. The dialogue is in Arabic script inside English prompts; the gloss under each prompt is our translation and was not sent to the model. The clips on this page are muted and compressed for the web.

- [Seedance 2.5: specs, inputs and credits](https://twistavo.com/models/seedance-2-5)
- [Storyboard sheet prompts](https://twistavo.com/prompts/storyboard-sheet)
- [Another talking animal: a dog hosts a talk show](https://twistavo.com/prompts/talking-dog-talk-show-ai-video)

## 1. The joke as a 6-panel board before any video

Result: AI storyboard made with Seedream 5.0 Pro: six photographic panels of one macaque in a teal sweatshirt in a kitchen, reading a blank note, holding a banana, cutting it, holding two halves, eating, and smiling over the peels

Image: https://twistavo.com/prompts/talking-animal-sitcom-ai-video/storyboard-board.webp

Settings: Made September 23, 2026, Text to image, Aspect ratio 3:4, Resolution 2k

```text
Create ONE production storyboard CONTACT SHEET made of SIX genuine cinematic photorealistic AI-generated images, NOT sketches, NOT vector art, NOT diagrams. Exactly six equal distinct panels in a strict 3-column by 2-row grid, left-to-right top row then left-to-right bottom row, thin straight dark gutters and nothing else. Each individual panel is composed like a vertical 9:16 smartphone film still. No decorative border, no header, no captions, no numbers, no Arabic or English words, no subtitles, no logos, no watermark, no text on props. This sheet will be cropped into six separate shot images. Do not reproduce any reference image or existing franchise character.

CHARACTER AND LOCATION LOCK: One and only one original adult macaque with short chestnut-brown realistic fur, a distinctive cream-tan oval muzzle, small round ears, two warm amber eyes, a short darker brown tuft between his brows, expressive but natural facial anatomy, and the SAME plain teal crew-neck sweatshirt in every panel. Same recognizable face, same fur pattern and garment throughout. Human-like hands with five fingers but believable primate anatomy. An ordinary warm-lit small kitchen in every panel: cream tile wall, walnut counter, wooden cutting board. Same daylight direction and photographic camera color; no other animal or person.

STORY BEATS, ONE CLEAR PHOTOGRAPHIC SHOT PER PANEL:
1. CLOSE-UP: the same macaque reads a small blank cream doctor prescription paper at the kitchen counter, puzzled but serious. Paper has no letters at all, only empty space. His face is visible beside the paper.
2. FACE CAMERA: the same macaque holds up ONE whole ripe yellow banana and gives the lens an unusually earnest look, as if relaying the doctor's instruction. Banana intact.
3. INSERT: on the wooden cutting board, the macaque's hands cut THAT ONE banana across the middle using a small table knife. The two cut ends are starting to separate. His teal sleeves and fur are the same as the other frames. No second banana.
4. MEDIUM SHOT: the macaque proudly holds the TWO distinct banana halves, one half in each hand, both cut ends clearly visible. No intact extra banana.
5. CLOSE SHOT: he has eaten one half and is bringing the remaining half to his mouth; one empty peel lies on the counter. Show the action unmistakably, no intact banana, no duplication.
6. CLOSE FACE-CAMERA SHOT: he has eaten both halves; he gives a cheeky knowing smirk straight into the lens. Exactly two empty peel pieces on the board, no banana flesh left.

These six frames are a coherent temporal sequence, with a natural visual punchline. No change of species, outfit, kitchen, banana count, camera grade or lighting. Avoid comical illustration style: photoreal tactile fur, fabric and kitchen materials, cinematic live-action-style stills, crisp expressions. No speech bubbles; spoken Algerian Darija and all on-screen text will be added separately in editing.
```

Tip: The "CHARACTER AND LOCATION LOCK" block held: one face, one teal sweatshirt and one kitchen in all six panels, and the note stayed blank. Panels 1 to 5 follow the beats. Panel 6 did not: it shows two banana pieces on the board where we asked for empty peels. We kept the board anyway and, in the Seedance prompt further down, told the model to use panel 6 for the face only. A board costs a small part of one video run (compare the costs under the examples), so it is the cheapest place to find a beat that will not read.

## 2. Gemini Omni Flash 1.1: the first part, voice generated by the model

Result: AI video made with Gemini Omni Flash 1.1: a macaque in a teal sweatshirt holds a banana to camera, cuts it on a wooden board, then bites one half while holding the other

Image: https://twistavo.com/prompts/talking-animal-sitcom-ai-video/gemini-kitchen-part-1.webp

Settings: Made September 23, 2026, Reference to video, 5 reference images (board panels 1 to 5), Aspect ratio 9:16, Duration 7 s, Resolution 720p, Native audio: the model voices the lines

```text
GEMINI OMNI FLASH 1.1 — PART 1/2, FIRST SEVEN SECONDS OF ONE ORIGINAL COMEDY VIDEO. Generate a MOVING 9:16 cinematic social-video scene, not a still-image montage, contact sheet, animal stock footage, silent video or frozen talking head. The five attached pictures are a chronological visual storyboard of ONE original chestnut macaque wearing the SAME plain teal knitted crew-neck shirt in the SAME warm small kitchen, not five different animals or images to pan across. @Image 1: his recognizable face, blank medical note and kitchen. @Image 2: the SAME macaque holds ONE whole yellow banana. @Image 3: his hands and knife make a crosswise cut through the banana on the wood board. @Image 4: TWO separate halves of that ONE banana, one in each hand. @Image 5: his face and motion while eating. Preserve the same cream-tan muzzle, little brown tuft, amber eyes, teal top, wood board, daylight direction and backdrop. Believable primate hands and natural blinking, chewing, breathing and fur motion, with real physical action; no new banana, person or room.

START at the blank note and quickly look directly at the viewer. 0–2.3 seconds: he lifts the one intact banana and says the first line while facing the camera. 2.3–5 seconds: he sets that banana on the board, peels it and makes ONE unmistakable crosswise knife cut, physically separating the fruit into exactly two pieces; he says the second line, sets the knife aside. 5–7 seconds: cut to a medium view, he proudly shows a half in each hand, then takes and chews one decisive bite from the FIRST half, swallowing that first half as a comedic jump cut. End this first video in a stable waist-up shot with the same macaque holding ONLY the second remaining half in his right hand, his left hand empty, as he glances toward it. Leave room for the action to continue immediately in part two. A single shared spatial axis and natural motivated cuts. Do not eat or drop the second half in this part.

AUDIO is native and audible. The SAME single MALE fictional macaque speaks only these two lines, in colloquial NORTHWEST ALGERIAN DARJA AS SPOKEN IN MAGHNIA ON THE OUJDA BORDER, with natural local Maghnia/Oujdi cadence and pronunciation, emphatically NOT Modern Standard Arabic, eastern Arabic, Cairo, Casablanca or a generic dubbed newsreader. Record the actual colloquial words below, NOT a translation, not an explanation, no added narration. Say, deadpan with a small pause: «الطبيب قالّي: نقص من البنان.» Then, while cutting: «قطّعتها على زوج...» Pronounce «زوج» in western Algerian border vernacular. Mouth movement synchronized on face shots. Small natural room tone and slicing sound only; no instrumental music, no laughter track. No generated text, subtitles, graphics, brands, watermark or doctor on-screen. This is part ONE only; do not give the last punchline yet.
```

Tip: The five panels worked as a shot list: lift the banana, cut on the board, a half in each hand, a bite. A run on this model is shorter than the whole joke, so we split it into two parts and gave each action a time range in the prompt. The part ends on a held pose with one half left, so a second clip can pick up from it. The cost of that split: two runs and an edit point, against one run on Seedance below.

## 3. Seedance 2.5: the whole joke in one run from the same board, lip-synced to a voice file

Result: AI video made with Seedance 2.5: a macaque in a teal sweatshirt raises a banana to camera, peels and cuts it in two, eats both halves and ends on a close-up smirk above the empty peels

Image: https://twistavo.com/prompts/talking-animal-sitcom-ai-video/seedance-kitchen-full.webp

Settings: Made September 23, 2026, Reference to video, 6 reference images (board panels 1 to 6) and 1 audio reference (the voice file), Aspect ratio 9:16, Duration 12 s, Resolution 480p, Audio on, bitrate high

```text
Create one finished 12-second vertical 9:16 live-action-style comedy VIDEO, not a contact sheet, slideshow, frozen photograph, or talking portrait. All 12 seconds have real continuous subject/action motion, natural blinks, breathing, fur and sleeve movement, hand motion, chewing, and visible mouth articulation timed to @Audio 1. One original adult chestnut macaque with the cream-tan muzzle, short dark tuft, amber eyes and plain teal knitted crew-neck top from the references, in the same warm kitchen at the same wooden cutting board. No other person or animal. Warm natural light, believable primate anatomy, close-to-medium edits, fast readable social-video rhythm; serious deadpan acting is the joke.

The six images are a temporal storyboard, not six static frames to display. @Image 1 supplies the puzzled opening reading a blank medical note; @Image 2 supplies the same macaque holding ONE whole banana to camera; @Image 3 supplies the tight insert of the cut; @Image 4 supplies the TWO pieces, one in each hand; @Image 5 supplies the eating action; @Image 6 is a FACE-ONLY identity/expression reference for the final knowing look, NOT a reference for banana pieces remaining on the table. Keep one consistent macaque, outfit, room, board and banana throughout; nothing should jump, multiply or change hands unnaturally.

0–3 seconds: he lowers the blank note, looks at the camera and raises his one whole banana as he delivers the doctor's instruction. 3–5.3 seconds: cut to his hands, he peels that ONE banana and makes ONE visible crosswise cut right through it into precisely two halves, placing the small knife safely aside, saying he cut it in two. 5.3–8.9 seconds: brief proud reveal with one half in each hand; then he visibly brings the first half to his mouth, bites, chews and swallows it, and does the same to the second half with comic haste. By 9 seconds both halves have been eaten, hands are empty, and only peel remains on the board. 9–12 seconds: face-camera close shot as he delivers the punchline with a small eyebrow lift, mouth moving naturally, then a brief cheeky silent look directly into lens. Do not leave any edible banana flesh in the final shot. Never cut to an extra fruit, another location, or a stationary hold instead of the physical act.

@Audio 1 is the authoritative 12-second original Algerian Darija MALE voice and cadence, with four phrases and planned pauses. Synchronize the visible speech and cutting/eating to this reference; do not change, translate, replace or repeat the dialogue: «الطبيب قالّي: نقص من البنان.» then «قطّعتها على زوج...» then «وكلّيت الزوج!» and finally «قالّي نقص، ما قالّيش حبّس.» Light kitchen room tone and small synchronized slicing/chewing sounds are fine; no music, no subtitles, no on-screen words, no logos or watermarks. The note remains blank: all exact subtitles will be added in editing, not generated into the picture.
```

Tip: One run told the whole joke: note, banana to camera, the cut, a half in each hand, both halves eaten, and a close-up for the punchline with only peel on the board. The line that fixed panel 6 was "FACE-ONLY identity/expression reference ... NOT a reference for banana pieces". With a voice file as @Audio 1 the timing comes from the recording, so write the action timings to match its pauses. Compared with Gemini: one run for the full joke at 480p, against a 720p first part with a second part still to make; the costs are under each example.

## 4. Two characters, two voices: the one clean take out of ten

Result: AI video made with Gemini Omni Flash 1.1: two macaques on a sofa in a lamp-lit living room, one in a cream roll-neck sweater and one in a plain navy hoodie who slides a phone away across the tea table, then both laugh

Image: https://twistavo.com/prompts/talking-animal-sitcom-ai-video/gemini-sofa-duo.webp

Settings: Made September 29, 2026, Reference to video, 1 reference image (the two characters), Aspect ratio 9:16, Duration 9 s, Resolution 720p, Native audio: the model voices both characters

```text
Original photorealistic Algerian comedy sketch with exactly TWO anthropomorphic macaques, zero humans. @Image1 defines these two existing characters: LEFT, stocky grey-brown macaque wearing a chunky cream roll-neck sweater; RIGHT, slimmer brown macaque with large ears and dark quiff wearing a navy zip hoodie and dark trousers. Preserve faces, fur, builds and outfit colours. Clothes plain; no emblem or visible writing. The reference is an identity bible, not a fixed pose. A modest warm Algerian living room, sofa and low tea table, evening lamp. Natural sitcom medium two-shot with one motivated closer reaction, no random montage. Spoken dialogue in informal western Algerian Darija (Maghnia/Tlemcen), not formal Arabic, not a narrator. LEFT has a lower calm voice, RIGHT a lighter cheeky voice; never swap speakers. Native room ambience, clear humanlike speech and naturally audible shared laughter. No music, subtitles, logos, television insert, applause or laugh track. FIRST HALF of one natural conversation. RIGHT keeps his plain phone face down and slides it farther away with one finger, avoiding it; LEFT laughs at the obvious self-deception. No bank name, balance, UI or real financial advice. In 0.5-8.5 seconds perform these three lines in order, with clear turn taking and no overlapping speech: LEFT says once: « راك توفّر الدراهم هاذ الشهر؟ ». RIGHT says once: « راني مرتاح بزاف! ». LEFT says once: « كيفاش؟ ». End on RIGHT preparing to answer the last question. No punchline or laughter yet. No added words.
```

Tip: Two speakers work when each gets a side (LEFT, RIGHT), a voice ("lower calm", "lighter cheeky") and one line at a time with "no overlapping speech". The phone stayed face down, as asked, so no screen had to be checked. This was the only take of the ten we ran that day with a plain hoodie: the reference image itself had a crest on the hoodie, and the other nine kept a logo or a badge. Clean the reference image first; the notes give the details.

## Notes

What went wrong: on September 29 we ran ten sofa takes with the same reference image and the same prompt opening, which says "Clothes plain; no emblem or visible writing". The reference image itself shows a football federation crest on the navy hoodie, and the model carried it over: the plain-clothes line gave a plain hoodie in only 1 of the 10 takes, the one shown above. In eight takes the hoodie had a sportswear logo, and in five of those also the crest. A ninth take gave the hoodie character a wristband with a printed badge. The lesson: clean the reference image before you generate, because a line in the prompt does not reliably remove what the reference shows. Then check frames before you pay for a second part. Every animal on this page is invented and every clip is AI-generated. The voice file of the Seedance example was made separately and is not priced here. Credit costs are Twistavo's quotes for the same model and settings on November 18, 2026. Results change from run to run.
