Key takeaways
- A 10-second, two-shot scene with dialogue cost 214 credits; the two 5-second 720p clips were 170 of it.
- Making the second frame as an edit of the first kept the same woman, coat and letter in both shots.
- Both lines were spoken word for word; we checked them with a transcription.
The kit, made in Claude

Vertical cinematic still from a short drama: a young woman in a rain-soaked trench coat stands at the door of a small night-time flower shop, holding an old letter, neon sign glowing pink behind her, tense and moody, film grain
GPT Image 2.6 Flare · 20 credits

Vertical cinematic still, same scene and same young woman in the trench coat holding the letter: now inside the flower shop, an older man in an apron looks up from arranging roses behind the counter, surprised, warm lamp light, film grain [first frame attached]
GPT Image 2.6 Flare · 20 credits
Shot 1: She looks down at the letter, then up at the shop, and says quietly: "He kept every one of them." Rain, distant traffic, a soft buzz from the neon sign. No music. Shot 2: The man puts down the roses and says, stunned: "I didn't think you'd ever come back." The shop bell rings softly. No music. Then join the shots and add clean captions.
2 × Gemini Omni Flash 720p, 5 s + assembly · 174 credits
Total: 214 credits for 3 outputs.
How to do it
- Write the scene: who, where, and one line per shot.
- Make the first frame, then the next frames as edits of it, so characters stay the same.
- Animate each frame with its spoken line in quotes and the ambient sound.
- Join the shots and caption them.
What to ask Claude
PromptWrite a 10-second short-drama scene in two shots: a woman returns to her estranged father’s flower shop with an old letter. One line each.
PromptMake a vertical cinematic still for shot 1, then shot 2 as an edit of it with the same woman.
PromptAnimate both at 720p with the lines spoken, join them and add clean captions. Price it first.
What it cost
| Piece | Model or tool | Credits |
|---|---|---|
| Two cinematic frames | GPT Image 2.6 Flare | 40 |
| Two 5-second clips with speech, 720p | Gemini Omni Flash | 170 |
| Join and caption | Merge media, Add captions | 4 |
| Total | 214 |
What we’d do differently
- The clean caption style holds the last words (“of them.”) through the pause before the next line. Fine here; for longer pauses, split the captions per shot.
- For a full episode, plan every shot first and draft at 360p; only render the takes you keep at 720p.
More on this: consistent characters and Claude video generator.
Frequently asked questions
Can AI make a short drama with dialogue?
Yes. Video models with native sound can speak a line you write in quotes; here both lines came out word for word.
How much does an AI short drama cost?
This 10-second scene cost 214 credits ($2.14). Drafting the shots at 360p first would cut the video cost by about two thirds.
Keep going
- Make a full ad campaign inside Claude: a real kit and what it cost
- UGC video ads inside Claude: a creator-style ad with speech and captions
- Faceless reels inside Claude: a 12-second fact reel with voiceover and captions
- AI real estate photos and video in Claude: virtual staging and a listing clip
- How to connect
- Pricing
Make this in Claude
Typeshoot adds image, video, voice and editing tools to Claude, ChatGPT and other assistants. 100 free credits to start, no card needed.
Get 100 free credits