Podcast presenter prompts: one host, one studio, every episode

What makes podcast prompts work

Fix two things: the host, and the studio itself. In RefVid the host is a saved character, and a studio photo from the main camera angle goes in as the scene photo. That keeps the set, mic and backdrop the same across episodes. Without it, the shelves, colors and lamp change every clip.

Frame like a podcast: seated, chest-up, a broadcast microphone on a boom arm in front of the host, slightly off-center. Name the camera as "wide" or "close" and keep those two framings for the whole series.

Keep each spoken line short, around 15 to 20 words for an 8-second clip. Long monologues are not something these models do yet; record the full episode for real and use AI clips for intros, promos and cutaways.

The prompts below give length, aspect ratio and sound in brackets; set those in the studio, not in the prompt text. With one host, the prompt says "the host" and RefVid attaches the saved character. With two, it names them: Ava and Jun are example names, so write your own saved names there.

Host intros

Prompt (8 s, 16:9, sound on): The host sits at the desk in the studio, a broadcast mic on a boom arm in front of him. He looks into the camera and says: "Welcome back. Today we are talking about why small teams ship faster." Close shot, chest-up, static camera. Warm key light from the left, soft blue backlight on the wall.

Why it works: a fixed close framing and a line under 20 words keep the delivery steady. Reuse the same prompt each week and change only the line.

Prompt (6 s, 16:9): Wide shot of the studio with Ava and Jun seated across a round table, each with a mic. They turn to each other and laugh at the same moment, then look toward the camera. Static camera, warm practical lamps, acoustic panels behind.

Why it works: a shared laugh is a natural opener with no dialogue to match. Save each host as a separate character and name both, so faces do not blend.

Prompt (6 s, 16:9, sound on): The host leans toward the mic, rests her elbows on the desk and says quietly: "This next part surprised me." Close shot, slow push-in, low warm light, background dark and soft.

Why it works: the lean and push-in create tension for a teaser. Use it as the cold open before the theme music.

Clips for social feeds

Prompt (5 s, 9:16): Close-up of the host reacting to something off-camera: eyebrows rise, she shakes her head slowly and smiles. Mic visible at the bottom of the frame, soft key light, studio blurred.

Why it works: reaction shots are among the most shared podcast moments and need no words. Pair one with a real audio clip in your editor.

Prompt (8 s, 9:16, sound on): Close-up of the host counting points on his fingers as he speaks: "One, write it down. Two, show someone. Three, ship it." Static camera, warm light, studio lamp in soft focus behind.

Why it works: a short list with gestures holds attention on mute. Keep finger counts to three; more than that tends to produce extra fingers.

Studio cutaways

Prompt (4 s, 16:9): Close-up of the host's hands sliding faders up on a small mixing desk in the studio, level meters glowing green. Shallow depth of field, warm lamp light.

Why it works: cutaways cover edits in the real recording. Keep hands simple and avoid readable labels on the desk.

Prompt (6 s, 16:9): Slow dolly shot across the empty studio before recording: two chairs, two mics on boom arms, a lamp turning on. In the last seconds the host walks in and sits down. Evening, warm and quiet.

Why it works: an empty-set opening works as an intro. In RefVid every video starts from a saved subject, so giving the host a late entrance is more reliable than asking for a room with nobody in it.

Spoken lines and length

In RefVid, turn sound on and the model, Kling 3.0 Omni, generates the voice and room sound with the picture. A line written in quotes is said as written; English and Chinese speech sound best. Check every take that the mouth matches the line; if it slips, shorten the line rather than rewriting the scene.

Clips run 3 to 15 seconds, and one clip can hold up to 3 shots, with the length split evenly between them. A 9-second clip could be shot 1, a wide of Ava and Jun; shot 2, a close-up of Ava as she speaks; shot 3, Jun's reaction.

On other models, lengths differ: as of October 2026, according to Google's Veo docs, Veo 3.1 makes 4-, 6- or 8-second clips, and 8 seconds when reference images are used.

Every video of the host is listed on its page in the RefVid studio, so a season of intros is easy to compare for drift.