
Elite Path
Vertical Video for Audio-Only Podcasts: Building Teasers From Generated B-Roll
Podcast discovery has drifted toward video. Short vertical clips on social feeds now do a lot of the work that word of mouth and directory rankings used to do, and video platforms have become a place where many listeners find new shows. That leaves audio-first podcasters in an awkward spot. If you record in a closet with a good mic and no camera, you have great audio and nothing to put on screen. The usual fixes are the static audiogram (cover art plus a waveform) or a talking head you never wanted to film. Neither stops a thumb mid-scroll. A third option has become practical: generate short, mood-matched b-roll clips with an AI video model and lay your real episode audio over them. Why b-roll works better than a waveform A waveform tells the viewer "this is audio". Moving imagery tells them what the story is about. A true crime show can open on a rain-streaked car window at night. A history podcast can open on candlelight over an old map. A cooking show can open on steam rising off a pan. The picture sets the tone in the first second, and your voice does the rest. The key point is that the generated video is supporting material. Your host's voice, the actual clip from the episode, is still the content. The visuals just give people a reason to stay long enough to hear it. Picking a model for the job Google's Veo 3.1 is a sensible fit for this because of its format. It makes clips of four, six or eight seconds, at up to 1080p, and it supports 9:16 vertical output natively, so there is no cropping a widescreen shot down for Reels, Shorts or TikTok. It also generates sound with the picture, which is usually a feature but needs handling here. On the Standard and Fast tiers you can switch audio off and get a silent clip, which is what you want under a voice track. The Lite tier always includes sound, so if you use it, mute the clip in your editor. Silent clips on Standard and Fast also cost less per second than clips with audio. The workflow, step by step 1. Pick the moment first. Find a 20 to 45 second stretch of the episode that works on its own: a strong opinion, a reveal, a funny exchange. Everything else follows from this. 2. Write down the mood in three words. "Tense, late, urban." "Warm, nostalgic, rural." These become your visual brief. 3. Plan three to five shots. At eight seconds each, four clips cover a 30 second teaser. Vary them: one wide establishing shot, one close detail, one slow camera move. 4. Write each prompt as a shot note. Shot size and camera move, one subject doing one thing, the light, the lens. Set the aspect ratio to 9:16. 5. Generate with audio off on Standard or Fast, or mute Lite clips afterward. 6. Edit. Lay the episode audio on the timeline, cut the clips to the rhythm of the speech, and add captions. Most viewers scroll with sound off, so burned-in captions of the actual quote matter more than anything else on screen. 7. Add the show name and episode number in your editor, not in the prompt, so the text is crisp and consistent across every teaser. Example prompts by genreKeep each prompt to one clear action. These are written for vertical b-roll with no people speaking: • True crime: "Vertical close-up, locked camera. Rain running down a car side window at night, blurred red and white city lights beyond, a single wiper sweep at the edge of frame. Moody, low contrast, 50mm lens, fine film grain." • History: "Vertical overhead shot, slow push in. A hand-drawn parchment map on a wooden desk lit by one flickering candle, a brass compass resting on the corner. Warm amber light, deep shadows." • Business or tech: "Vertical medium shot, slow pan. An empty open-plan office at dawn, monitors glowing softly, long shadows across the carpet. Cool blue light, calm and quiet." If you want to see what fuller prompts look like before writing your own, pixeldojo.ai publishes the exact prompt behind each Veo 3.1 sample clip along with what each tier costs, which is a quick way to learn the structure. Keeping it honest and on-brand • Don't fake the guest. Never generate a clip that implies a real guest was filmed. Stick to atmosphere, places and objects. • Keep a visual style. Pick a lens and color note, such as "35mm, warm grade, film grain", and reuse it in every prompt. Over a season, that becomes part of the show's identity. • Disclose where it's expected. Some platforms ask creators to label realistic AI-generated video. Check the current rules for each one you post to. • Batch it. Generate the b-roll for several episodes in one session once you know your style. Many clips, especially places and textures, can be reused across episodes. What to measure Treat the first few teasers as an experiment. Compare them with your old audiograms on the numbers each platform gives you: watch time, completion rate, and profile visits or follows. If the generated b-roll holds attention longer, build it into your production routine. If it doesn't, the issue is usually the moment you chose rather than the visuals, so go back to step one and pick a stronger clip.

