Image to Voice — Turn a Character Image into a Speaking Voice
No reference recording, no voice actor. Upload a single character image — a portrait, an illustration, a game character sheet — and Seed Audio 1.0 infers a voice that fits the face, then performs your script with it inside a complete audio scene.
Give a face a voiceWhat image-to-voice unlocks
Hear inferred voices in full scenes
Frequently asked questions
You upload one character image alongside your prompt. Seed Audio 1.0 analyzes the character — apparent age, mood, design — and infers a voice profile that fits, then uses that voice to perform the dialogue in your generated scene.
One. A single generation accepts one character image for voice inference. If your scene has several characters, the image-cast character can still interact with other prompt-described voices in the same mix.
No. Image inference and @Audio reference cloning are mutually exclusive per generation. Choose the image path when the character has no existing voice, or the audio path when you need to match a specific real voice.
A clear, front-facing depiction of a single character — photo, illustration, 3D render, or character sheet. The more legible the character design (age, expression, style), the more the inferred voice will feel like a natural fit.
No. Image to voice infers a plausible voice that matches how the character looks — it does not know or reproduce the actual voice of a person in a photo. To match a specific real voice, use reference audio cloning instead.
Yes. Only upload images you have rights to, and get consent before building voiced content around an identifiable real person. For fictional characters and your own artwork, you are free to create.
Within a scene and its Extend continuations, yes — Extend keeps character voices consistent as the story grows past the ~2-minute single-generation limit.
Voiced comic dubs and webtoon recaps, game NPC dialogue, VTuber and virtual-idol content, animated shorts, brand mascot spots, audiobook character casts — any project where a designed character needs to speak.
Yes — signing up grants free credits, and image-based generations spend the same credits as any other scene. Generated audio can be downloaded and used commercially.
Each work saves to your library with a public /audio/{id} page. Send that link to your team, your artist, or your community to let them hear the character before you publish.