A manga panel already has everything a scene needs: characters, a setting, and dialogue in speech bubbles. What it does not have is sound. Turning that dialogue into spoken lines is what makes a still panel feel like a moment from an anime.
This guide focuses on the voice side of making a motion comic: how the dialogue in your panel becomes speech, how to give each character the right voice, and what to do when a voice comes out wrong.
Key takeaways
- The dialogue already written in the speech bubbles becomes the spoken lines. You do not need to retype it.
- Add voices from the Voice Library or upload your own with Use Local Voice, then type @ in the prompt to place each voice next to its character.
- Give each character one voice and describe them the way they look in the panel, so the right voice goes to the right character.
- Keep the speech bubble content unchanged, so the spoken lines and the text on screen match.
- Panels without dialogue still work. Ask for ambience, sound effects, or music instead of speech.
How does dialogue in a manga panel become speech?
When you animate a panel in Comik AI, you can ask the AI to generate character voices based on the dialogue in the comic. It reads the text in the speech bubbles and turns each line into speech for the character who says it.
That means the panel itself is your script. Write the dialogue in the bubbles the way you want it spoken, then tell the AI which voice belongs to which character.
What do you need before you start?
- A manga or comic panel, or a full page, with clear speech bubbles. Short, readable lines work best.
- Art you own or have permission to use. Use your own work or images you created with Comik AI.
- A rough idea of each character's voice. For example, a calm young woman or a cheerful child.
How do you add voices to a manga panel in Comik AI?
Open the AI Motion Comic Generator and follow these steps.
Step 1: Upload the panel
Click References and upload your manga panel, or a full page with several panels. Keep the default mode, Reference to video. It accepts audio references, which you need for character voices.
Step 2: Choose an animated style
Pick Original Art to keep your art exactly as drawn, or Auto Color to add color to black-and-white manga as it moves. For a full page, choose Multi-Panel Original Art or Multi-Panel Auto Color. The voice setup is the same for all four.
Step 3: Start from the default prompt
While the prompt box is empty, click Use Default Prompt to fill in a starting prompt. This is the default prompt built into Comik AI. It is written for a panel with one male and one female character, and gives each one a voice:
Default Prompt
Convert this comic image into an anime-style animated video.
Keep the characters, scene, composition, camera angle, clothing, props, expressions, and speech bubble content unchanged. Do not add, remove, replace, or redesign any elements.
Generate character voices based on the dialogue in the comic:
- Male character voice: [@Broadcast Host]
- Female character voice: [@Sweet Girl]Edit the voice lines to match the characters in your own panel.
Step 4: Assign a voice to each character
First, add the voices you want to use. Choose them from the Voice Library, or upload your own recording with Use Local Voice. Then type @ in the prompt and pick a voice for each character. The voice is inserted as a mention, like [@Broadcast Host]. Keep one line per character, with a short description in front of the mention.
Step 5: Generate and listen
Before you generate, check that Output Audio is set to With audio. A video generated without audio has no voices. Then generate the video and watch it with the sound on. Check that every line is spoken, that each line comes from the right character, and that the voices fit the characters.
How do you choose the right voice for each character?
- Match the voice to how the character looks. Age, energy, and personality matter more than anything else. A soft-spoken voice suits a quiet character; a bright voice suits an energetic one.
- Describe the character the way they appear in the panel. "Young woman voice" or "Old man voice" is easier for the AI to match than a name it cannot see.
- Match the voice language to the dialogue. The Voice Library can be filtered by language and accent, so if your speech bubbles are in Japanese, start with Japanese voices.
- Use one voice per character. Reusing the same voice for two characters makes it hard to tell who is speaking.
- Keep the same voice across panels. If a character appears in several shots, use the same voice every time so they sound like the same person. This matters most with a Multi-Panel style, where one character can speak in several scenes of the same video.
What if a panel has more than two characters?
Add one line per speaking character, each with its own voice. Make the descriptions easy to tell apart, for example by age, role, or position in the panel: "Girl on the left voice" and "Boy on the right voice."
If a panel is crowded with dialogue, consider splitting the scene into separate panels. Fewer lines per shot are easier to voice correctly and easier for viewers to follow.
What about panels without dialogue?
Not every panel needs speech. For a quiet or action-only panel, leave out the voice lines and describe the sound you want instead, such as rain on the window, footsteps, or soft background music. A silent beat between two dialogue panels often makes the spoken lines land harder.
What are common voice problems, and how do you fix them?
A line is spoken by the wrong character
Make each description match how that character appears, and keep exactly one voice mention per character.
A line is missing or changed
Keep "speech bubble content unchanged" in the prompt, and make sure the bubble text is clear and readable in the panel. Very small or stylized lettering is harder to read.
The voice does not fit the character
Try another voice for that character and generate again. Changing one voice at a time makes it easy to see what improved.
New text appears on screen
Keep "Do not add, remove, replace, or redesign any elements" in the prompt, and do not ask for captions that are not in the original panel.
Lines sound rushed or get cut off
Choose a longer video length. About 6 seconds is usually enough for a single panel with two speech bubbles, while a four-panel page with eight speech bubbles needs at least 20 seconds, ideally 25. The maximum length depends on the model: up to 30 seconds with Seedance 2.5, 15 seconds with Seedance 2.0 and MiniMax H3, and 8 seconds with VEO 3.1.
Frequently asked questions
Do I need to type out the dialogue?
No. The dialogue in the speech bubbles becomes the spoken lines. You only assign a voice to each character.
Can different characters have different voices?
Yes. Add a voice for each character, then type @ in the prompt to place each voice next to the character it belongs to.
Can I use my own voice recording?
Yes. Upload your own audio with Use Local Voice, then mention it with @ like any voice from the Voice Library.
Can animals or non-human characters have voices?
Yes. Describe the character, such as "Cat voice," and assign a voice to it, the same way the default prompt does for its male and female characters.
Can I add sound effects and music too?
Yes. You can add sound effects and background music to match the scene, alongside the character voices.
Does voiced dialogue work with every animated style?
Yes. The voice setup is the same for Original Art, Auto Color, and both Multi-Panel styles.
Give your manga a voice
Pick a panel with clear dialogue, start from the default prompt, and assign a voice to each character. Open the AI Motion Comic Generator to make your first voiced motion comic.
