Talking dog talk-show video with AI: photo, set, voice and the episode

Seedance 2.5 and GPT Image 2 prompt on Twistavo, an AI video studio for character swap: replace a person in a real video from 1 to 4 photos; plans from $3.99/month.

Updated

The format: your dog is the host of a late-night talk show and delivers one short joke to camera. We built one episode's pieces in a single afternoon (September 17, 2026) for an invented golden retriever called Bruno: a set of photos that fix the dog, a stage key art, a storyboard for the opening camera move, the opening shot itself, a voice line and the talking clip. Each piece is below in the order we made it, with the exact prompt, the model, the settings and what it costs in Twistavo today. If you have a real dog, step 1 is a photo you take yourself, and the rest is the same.

Make your own with Twistavo — 1 free image

Setup

Model
Seedance 2.5 and GPT Image 2
Settings
Images with GPT Image 2 at quality high and 2k; video with Seedance 2.5 from one image, sound off. The voice line was generated separately as a 24 kHz WAV with Seed Audio from this text: "Hey, it's Bruno. Welcome to the show. First question: do I stare at the wall? No. I meditate. That wall owes me money." It was then given to Seedance 2.5, as an MP3 file, as the audio input of the talking clip. The talking clip comes back without a sound track (sound was off), so the voice file goes back under the picture in your editor. The pieces add up to one episode: the push-in opens the show, the talking clip carries the joke. The clips here are muted and compressed for the web.
Models used
GPT Image 2, Seedance 2.5

1. The dog: a phone-style photo that fixes the look

AI photo made with GPT Image 2: close-up of a cream-golden retriever with warm brown eyes and a big open-mouth grin in a backyard at golden hour
  • Made September 17, 2026
  • Text to image
  • Aspect ratio 4:5
  • Quality high
  • Resolution 2k

61 credits in Twistavo

Prompt

Photorealistic casual iPhone-style close-up selfie angle: Bruno, a 4-year-old cream-golden retriever male dog with a soft fluffy coat, warm brown eyes, a small white patch on the chest and slightly floppy ears, face filling the frame, big confident grin, outdoors in a backyard with soft bokeh, natural golden-hour light. No text, no watermark, no logo.

Tip: The prompt opens with a fixed description block (name, age, coat, eye color, chest patch, ears) that we pasted unchanged into three photos: this close-up, one on the couch and one in the kitchen. The three dogs read as the same dog. Writing "iPhone-style" and the owner's camera height keeps it from looking like a studio portrait, which matters later when the dog has to look like a pet and not a stock model.

2. The show: stage key art in clay

AI image made with GPT Image 2: a clay-style golden retriever in a dark suit and red tie grins behind a curved wooden desk at a vintage chrome microphone, a spotlight on the left and red curtains on the right
  • Made September 17, 2026
  • Image to image, 2 reference images
  • Aspect ratio 3:2
  • Quality high
  • Resolution 2k

61 credits in Twistavo

Prompt

Glossy 3D claymation diorama of a late-night talk show set: a golden retriever host sits behind a curved desk at a vintage chrome microphone, leaning in confidently with a sly grin, one paw on the desk. Around it: a small standing spotlight casting a warm cone of light, crimson curtains framing the right edge. Composition reserves open negative space on the left half for the type block (dark empty stage wall). Palette: deep charcoal #17161A stage, bone #F6F1E8 highlights, crimson #E63946 curtain accents, warm tungsten light. Cinematic lighting, shallow depth of field, premium playful 3D render. No text, no letters, no logos, no watermark. Full-bleed 3:2 landscape composition, edge-to-edge, filling the entire canvas — NO border, NO frame, NO rounded capsule, NO corner dots. Punchy contrast, committed saturated palette, commercial editorial grade. IMPORTANT: absolutely no text, no letters, no logos, no captions, no UI, no watermarks anywhere in the image.

Tip: The key art for the show page and the thumbnail. Hex colors in the prompt (#17161A, #F6F1E8, #E63946) held, and the left half stayed empty for a title as asked. What we did not ask for: the suit, the tie and the coffee mug. The model added them on its own; they fit the talk-show idea, but write "no clothes" if your dog must stay natural. Two uploaded style references went with this run; they are not shown.

3. The opening move: a 6-panel storyboard

AI storyboard made with GPT Image 2: six panels of one camera push-in on a dark talk-show stage, from an empty stage with a chrome microphone to a golden retriever walking in, sitting at the mic and grinning in extreme close-up
  • Made September 17, 2026
  • Text to image
  • Aspect ratio 16:9
  • Quality high
  • Resolution 2k

61 credits in Twistavo

Prompt

Storyboard for a 16:9 cinematic brand film, exactly six panels arranged in a 3x2 grid, each panel showing the SAME single continuous slow camera push-in, NOT six different scenes, no panel borders text, no letters anywhere. Subject: a photoreal golden retriever as the celebrity guest on a late-night talk show stage. Panel 1: very wide establishing shot of a dark elegant studio stage, charcoal backdrop, warm golden spotlight glow, a vintage chrome microphone on a stand, empty stage. Panel 2: slightly closer, the golden retriever walks in from the left toward the microphone. Panel 3: medium shot, dog sits at the mic stand, head alert, warm stage light. Panel 4: closer medium, dog tilts its head as if speaking, confident amused expression. Panel 5: close-up of the dog's face, microphone blurred in foreground, spotlight on fur. Panel 6: final extreme close-up, dog looks directly at the camera with a sly talk-show host grin. Consistent look: deep charcoal stage #141416, warm tungsten spotlight, bone-white highlights, cinematic grade, locked exposure, shallow depth of field, premium film photography. No text, no letters, no logos, no watermarks, no captions.

Tip: One image that shows the whole camera move as six moments of the same take, wide to extreme close-up. It costs a small fraction of the 1080p video it plans: compare the cost under this example with the next one. The prompt says "the SAME single continuous slow camera push-in, NOT six different scenes", and the panels kept one stage, one light and one dog. Check the board before you pay for the video: if the panels drift, the video will too.

4. The opening shot: a slow push-in to the host

  • Made September 17, 2026
  • Image to video, 1 reference image
  • Aspect ratio 16:9
  • Duration 15 s
  • Resolution 1080p
  • Audio off

2058 credits in Twistavo

Prompt

A single continuous slow cinematic camera push-in, one unbroken take, no cuts, no camera shake, no zoom pulses: a photoreal golden retriever sits at a vintage chrome microphone on a deep charcoal late-night talk-show stage. Camera starts very wide: the dog small on a dark stage with one warm tungsten spotlight, crimson curtain bokeh in the far background. Camera glides forward steadily, staying centered on the dog, the spotlight warming as we approach: medium shot of the dog alert at the mic, then close-up: the dog tilts its head slightly and gives a sly amused talk-show grin directly at the camera. Resolves on an extreme close-up of the dog's face, warm spotlight on the fur, shallow depth of field, dark seamless charcoal background behind. Deep charcoal stage #17161A, bone-white rim light #F6F1E8, crimson curtain accent #E63946, warm tungsten glow, locked exposure and white balance, consistent grade, minimal motion blur, premium film look, photorealistic. No text, no letters, no logos, no watermark, no captions.

Tip: One continuous push-in from wide to extreme close-up, as on the board, with exposure and color that do not jump. The prompt repeats "no cuts, no camera shake, no zoom pulses" and "locked exposure and white balance", which is what keeps a long move smooth. What to change: this is by far the most expensive piece. The same shot at 720p or 480p costs far less (check the quote in Create before you run it), and for a vertical feed the shot is cropped anyway.

5. The joke: the dog says the voice line

  • Made September 17, 2026
  • Image to video, 1 reference image
  • Audio input: the voice line (MP3)
  • Aspect ratio 9:16
  • Duration 8 s
  • Resolution 720p
  • Audio off

481 credits in Twistavo

Prompt

The cream-golden retriever dog named Bruno from the reference photo talking directly to camera as a comedic late-night talk show host, delivering the exact voice lines from the audio with moving mouth, lips synced to the speech, confident cheeky expressions, subtle head tilts, eyebrows raising, close to medium shot, sitting on the couch in the living room from the reference image, natural daylight, photorealistic, cinematic depth of field. Continuous single take, no cuts, no camera shake, no text, no watermark.

Tip: The reference is a photo of the same dog on a couch, and the voice file drives the mouth. The dog stays the same animal, on the same couch, in one take. What failed: the voice file runs well past the duration of the clip, so the end of the line, the punchline, falls outside the clip. Set the duration to the length of the voice (Seedance 2.5 accepts longer clips) or cut the line to fit. The clip also has no sound of its own: line the voice up in your editor and check the first syllable.

Make it in Twistavo

  1. Take or make one sharp photo of your dog looking at the camera, face large in the frame.
  2. Write the line in your dog's voice: under 10 seconds, one setup and one punchline. Generate it with a voice tool or record it yourself.
  3. Open Create in Twistavo, choose Seedance 2.5, add the photo as the reference and the voice file as the audio input.
  4. Paste the prompt of step 5, set 9:16, a duration at least as long as the voice file, 720p and sound off, check the credit quote and run it.
  5. Optional: make the stage key art and an opening push-in with the prompts of steps 2 to 4 for a show intro.
  6. In your editor, put the voice back under the talking clip and open the post on the dog already talking: the first two seconds decide whether people stay.

Variations to try

  • A cat as the grumpy late-night guest instead of the host, same couch prompt, a slower voice.
  • A weekly series: same set, same dog, a new short question each episode.
  • Two pets as co-hosts: make one photo with both and write a line for each.
  • Swap the talk show for a cooking show or a weather report and keep the push-in board as the intro.

Notes

Bruno is an invented dog: every image and clip on this page is AI-generated. A real pet photo works the same way as the input of step 5. The voice model (Seed Audio) is not priced in Twistavo's AI Studio quotes, so its cost is not listed; you can also record the line yourself. Credit costs are Twistavo's quotes for the same model and settings on November 18, 2026. Results change from run to run.

Make your own with Twistavo — 1 free image

More AI video prompts

This prompt, its model, settings and result are part of the open dataset of every tested prompt (CSV and JSON).