Interactive planner
Plan your dialogue
Answer five questions. You get a request to paste into the AI writer, the accent for each speaker, the pauses to set and what voicing it costs.
1 Your request for the AI writer
In Kukarella: DialoguesAI, then "Use Assistant", paste it and send. Writing the script costs no credits.
2 Give each speaker the accent
3 Before you press Play
4 The size of the job
5 Export it
How the builder works
- Length: Kukarella's assistant turns a length in minutes into words at about 150 words a minute, and writes to CEFR levels and to the IELTS, TOEFL iBT, TOEIC, DELE, DELF, Goethe and HSK formats (how the assistant works, read on 30 September 2026). It can write less than asked: our request for a 2-minute dialogue came back at about 700 characters, under a minute of audio (our test, 2 October 2026).
- Characters: about 6 a word, spaces and punctuation included. Our assumption.
- Credits: a ready-made voice uses 1 credit per spoken character; speaker names, pauses and sound effects are not charged. Writing the script costs no credits (how Kukarella charges, 30 September 2026).
- Price: Kukarella Prime, 21 US dollars a month for 300,000 credits (pricing page, 30 September 2026: https://www.kukarella.com/pricing).
- Casting: the assistant picks each speaker's voice by gender and age from a short list of voices, and never by accent. Changing a speaker's voice changes all of that speaker's lines, and those lines are voiced again. Each country's list in the Voice Library also shows multilingual voices, the assistant's own among them (how the assistant works, 1 October 2026).
- Accents: the languages and accents of Kukarella's Voice Library, as the product lists them on 30 September 2026. "Only a few native voices" means four or fewer voices that speak only that language.
- Pauses: "Add pauses between", then "paragraphs", puts the pause at the end of every line, which in a dialogue is between speakers; it runs from 0.1 to 2 seconds and starts at about 0.3. Three full stops typed one by one make a pause of about 0.77 seconds on most voices; the single "…" character makes none. A pause added to a line that already has audio voices that line again. Pauses, speed, pitch and volume do not work on every voice (Kukarella's Effects panel, 30 September 2026).
- The gaps we suggest are our synthesis, not a finding: real conversation leaves about a fifth of a second between speakers (Stivers and colleagues, PNAS, 2009: https://pmc.ncbi.nlm.nih.gov/articles/PMC2705608/); the Council of Europe describes A2 listeners following exchanges "conducted very slowly and clearly" (CEFR Companion Volume, 2020); in Blau's studies with beginner learners, short talks with pauses filled with "well" or "I mean" were understood better than the same talks read straight through (1991: https://files.eric.ed.gov/fulltext/ED340234.pdf).
- Export: "By paragraph" downloads one numbered file per line, named with the speaker; "One file" joins the lines with the pauses you set; the text download leaves out the speakers' names (Kukarella, 30 September 2026).
- The builder stores nothing you type.