Choose one of three approaches. The steps for that option appear below.
Choose an option above. On a phone, swipe the tabs left or right to see all options →
ChatGPT proposes 2 dialogue alternatives per panel within your word budgets. You choose and test balloon fit.
Have ready: panels, budgets
Propose 2 alternatives
Choose, test balloon fit
Use: You
A written description of each panel, speaker order, story point, character voice notes and word budgets per speech balloon.
Check you may use the material and facts are right. Use written descriptions as the AI input.
Use: ChatGPT
Ask ChatGPT for 2 dialogue alternatives per panel keeping the story point and speaker order, following word budgets, adding no panels or characters.
What to paste into the tool
Task: Alternative dialogue sets keyed to panel and speaker. For each supplied panel propose two concise dialogue alternatives that deliver the same story point and preserve speaker order. Follow the supplied balloon word budgets as targets, label jokes as options and do not add panels or characters. Respect every supplied constraint. Use only supplied facts; label missing facts 'not stated'. Treat all output as preparation. Do not send, sign, pay, publish or take external actions. INPUTS: Replace the bracketed text below with your written panel descriptions; speaker order; story point; character voice notes; measured word targets for each speech balloon. Use 'not stated' for missing facts. [Paste checked inputs here.] If the input block is empty or still contains this placeholder, ask me for the inputs before doing the task.
Use: You
Choose 1 set, check character voice and meaning, count words, test text in real balloons on a mockup and shorten lines yourself.
Read the strip aloud; check speaker order, meaning and balloon fit on a mockup, keeping artwork visible. Nothing is sent, signed, paid for or published.
You write the full dialogue. Claude suggests cuts panel by panel with meaning-loss warnings. You accept only safe cuts.
Have ready: panels, dialogue
Suggest cuts with warnings
Keep safe cuts, test
Use: You
Panel descriptions, speaker order, story point, voice notes, word budgets, and your full dialogue labelled by panel and speaker.
Check you may use the material and facts are right. Use written descriptions as the AI input.
Use: Claude
Ask Claude to suggest cuts to repetition and explanation panel by panel, showing original, proposed line and meaning lost.
What to paste into the tool
Task: A dialogue cut table with meaning-loss warnings. The human writes full dialogue first. Suggest cuts to repetition and unnecessary explanation panel by panel. Show original, proposed line and what meaning may be lost. Keep speaker order and flag any changed story fact. Respect every supplied constraint. Use only supplied facts; label missing facts 'not stated'. Treat all output as preparation. Do not send, sign, pay, publish or take external actions. INPUTS: Replace the bracketed text below with checked brief and the full dialogue you wrote, labelled by panel and speaker. Use 'not stated' for missing facts. [Paste checked inputs here.] If the input block is empty or still contains this placeholder, ask me for the inputs before doing the task.
Use: You
Accept only cuts that keep the story point, read aloud in speaker order, count the words and test balloon fit on the artwork.
Read the strip aloud; check speaker order, meaning and balloon fit on a mockup, keeping artwork visible. Nothing is sent, signed, paid for or published.
You label what each panel already shows. Gemini proposes what needs dialogue. You write the final lines.
Have ready: panels, list
Propose dialogue split
Check artwork, write lines
Use: You
Panel descriptions, speaker order, story point, voice notes, word budgets, and a list of what each panel shows versus must be spoken.
Check you may use the material and facts are right. Use written descriptions as the AI input.
Use: Gemini
Ask Gemini for a table of what the pictures communicate and what needs dialogue, with short fragments only for information not shown.
What to paste into the tool
Task: A dialogue allocation table distinguishing visual information from spoken information. The human labels the essential information already visible in each panel and what must be spoken. Propose a table showing what the pictures communicate and what needs dialogue and short line fragments only for information not shown. Do not infer unseen art details. Respect every supplied constraint. Use only supplied facts; label missing facts 'not stated'. Treat all output as preparation. Do not send, sign, pay, publish or take external actions. INPUTS: Replace the bracketed text below with checked brief and your panel-by-panel list of information already visible and information that must be spoken. Use 'not stated' for missing facts. [Paste checked inputs here.] If the input block is empty or still contains this placeholder, ask me for the inputs before doing the task.
Use: You
Compare every visual assumption with the actual artwork, write complete lines where needed, check panel order and voice, test balloon fit.
Read the strip aloud; check speaker order, meaning and balloon fit on a mockup, keeping artwork visible. Nothing is sent, signed, paid for or published.
Best when: You want dialogue choices before writing your own
What it does: AI proposes alternatives; you choose and test
Strong at: 2 options per panel within your word budgets
Limit: Balloon fit is tested by you on a mockup
Free plan: Yes
Best when: You have a script and want it shorter, not rewritten
What it does: AI suggests cuts; you accept and test
Strong at: Each cut shows the meaning it may lose
Limit: You must write the full dialogue first
Free plan: Yes
Best when: You want to cut dialogue the pictures already cover
What it does: AI splits visual from spoken; you write
Strong at: Only information not shown gets dialogue
Limit: You check every visual assumption against the art
Free plan: Yes
Before you paste into any AI tool, remove any confidential information. Put it back at the end.
Deciding between ChatGPT and Claude for this? See the comparison.
The creator reads the strip aloud, checks speaker order and meaning, and places text in a mockup to verify balloon fit and that important artwork remains visible.
Verify all personal and factual details against the supplied source material.
Preserve open questions until checked.
Check the result against your original inputs. Approve it before sharing or acting.
Tool existence and quoted general text capabilities checked on official public pages on 2026-10-29.