Build an AI micro drama series in Venice Studio

Why micro dramas and the four-step workflow

AI micro dramas are short, soap-opera-style vertical videos, and platforms like TikTok have begun testing feeds built largely from AI-generated content. This guide walks through making episode one of your own series entirely inside Venice.

The workflow has four stages. First, write a treatment and script with Venice chat. Second, generate character reference images. Third, render video clips using those references. Fourth, stitch the clips together and add a score. Everything happens in the Venice Studio, so you never switch tools.

Writing the treatment, script, and shot list

Open a new chat and ask it to build the foundation. The example prompt:

Let's create a treatment for a micro drama series and the script and shot list for episode 1.

Then feed it your ideas. Because Seedance 2.0 cannot use realistic human face references, steer toward a highly stylized look. The video uses an anime aesthetic with a handsome man and beautiful woman as protagonists, a dirty secret, set on a Colombian coffee finca with the vibe of a LatAm telenovela.

The genre matters. A micro drama is a soap opera, so push for extreme emotion and high stakes, because that is what hooks viewers. In the example, the model (GLM 5.1) named the series Cafe El Pecado, set an aesthetic of high-contrast neon noir against lush greenery, capped episodes at two minutes, and built the secret around a vanished abusive husband. Once you have the treatment and script, ask the chat for character reference image prompts and per-scene shot prompts so you can generate visuals deliberately instead of blind.

Generating character references in the Studio

Reference images exist so you can lock in what each character looks like before producing any video. Without them, every new shot drifts and you lose consistency.

Copy each character reference prompt from your chat, one at a time, and paste it into the image generator in Venice Studio. Set the aspect ratio to 9:16 for vertical video. The walkthrough uses Nano Banana 2, a top-tier model that costs around 10 credits per image. If you would rather not spend credits, a Venice Pro membership gives access to free image models, though quality will vary.

While one image renders, jump back to your chat, copy the next prompt, and queue it up. In the example this produced Isabella, the anime Latina protagonist, and Mateo, the groundskeeper.

Setting up Seedance 2.0 and tagging characters

Go to the video tab in Venice Studio and select Seedance 2.0 R2V (reference to video). Add your character reference images; they appear as image one, image two, and so on, which you can then reference directly inside prompts.

Paste your scene shot prompts and clean them up, tagging the correct reference image wherever a specific character appears. Tagging is what preserves details. In the example, the prompt calls for a close-up of "image one's calloused rugged hands" specifically to keep his wrist tattoo visible across shots:

Wide drone shot, anime style, Studio Ghibli painted background, sweeping vertical shot of a lush green coffee plantation.
Close-up of image one's calloused rugged hands.
Cut to medium shot of the handsome rugged groundskeeper, image one, wearing a straw hat, looking down at his hands.

One practical tip: remove a character reference from a shot if that character does not appear in it. Including an untagged character can confuse the model.

The no-music rule and multi-shot generation

End every video prompt with an instruction like "no music, only sound effects." You want the sound effects, but you will add a custom soundtrack later when stitching clips together. If a clip arrives with its own music, it conflicts with your edit and creates choppy cuts. When that happens, use the clip's menu to reload the source, append the no-music instruction, and regenerate.

Seedance 2.0 can render several cuts in a single clip. The example packs three shots, an establishing drone shot, a close-up of hands, and a medium shot, into one generation. Open the settings icon and give it enough time for all the cuts to play out: the video uses 12 to 15 seconds depending on how many shots are bundled. Always confirm the aspect ratio is 9:16 before sending, since forgetting it forces a re-prompt.

Parallel rendering, editing, and scoring

Video generation runs in the background, so while clips render you can return to your treatment chat and pull the next batch of prompts. This parallel approach keeps you productive instead of waiting.

When clips are ready, open the movie editor tab. Set the project aspect ratio to TikTok 9:16 so vertical clips fit cleanly, then insert clips one by one and sequence them. Make quick cuts to trim awkward sections, use fade-outs (around one to one and a half seconds) as transitions, and add transitions like a blur where you need a mood shift. Editing improves with each project, so treat rough early results as practice.

For the score, open the audio tab (the guide suggests using a second browser tab so you can switch back and forth). Note the timestamps where the mood changes, then describe them in your prompt. Two approaches work: write a manual timed prompt, or ask your original chat to generate a soundtrack prompt, since it already knows the story, scenes, and series. Example manual prompt:

Create a cinematic soundtrack for a 61 second micro drama. The first 14 seconds are exposition, calm. From 14 to 42 seconds it builds between the love interests and their mysterious dynamic. At 42 seconds a sudden intense sting comes in as he grabs her hand. From 45 seconds onward, mysterious cliffhanger.

In the audio settings, force instrumental so there are no vocals, choose an instrument set such as orchestra, and set a length that leaves a couple of seconds to fade out. Using the 11 Labs music model, generate both versions, compare them in the editor, and keep the better one. Add a title card with a fade, polish your transitions, and you have episode one. Continue the same chat conversation to produce the next episodes.

Key takeaways

• Build everything in order: treatment and script first, then character references, then video, then edit and score, all inside Venice Studio.
• Use a stylized anime aesthetic because Seedance 2.0 reference-to-video cannot use realistic human faces.
• Tag reference images (image one, image two) in every prompt to hold characters and details consistent across shots.
• End video prompts with "no music, only sound effects" so your clips stay editable and your own score lands cleanly.