年間会員 50%オフ - 期間限定！
Upload a portrait and a song or a voice recording. The AI animates the face with lip sync and natural head movement, and returns a singing video you can download.
Preset: 480p resolution. Upload a portrait and an audio file, optionally describe the performance, and generate.
This page uses InfiniteTalk, an audio-driven video model. Given a single still image and an audio track, it generates a video in which the person in the photo speaks or sings along, with lip movement matched to the sound and subtle motion in the head and expression.
It suits music projects: a singer's headshot becomes a performance clip, a character portrait becomes a virtual artist, and a family photo turns into a birthday greeting with a song.
Pick a sharp portrait with the face centred, eyes open and the mouth not covered. Avoid heavy filters, sunglasses and extreme side angles. A 1:1 or 4:5 crop with some space around the head gives the model room for natural movement.
For music, a clear vocal gives the best sync. If your track has a loud band, extract the vocal first and upload that. The AI lyrics and music tools can create an original song for the photo if you do not have one.
Billing follows the audio duration, rounded up in 5-second blocks, at 15 credits per block for 480p and 30 for 720p. A three-minute song is 36 blocks, so 540 credits at 480p. Test with a short excerpt first, then render the full track.
The model supports audio up to 10 minutes. For social posts, a 20 to 40 second chorus at 720p is often the sweet spot between cost and quality.
Lip sync quality depends on the clarity of the audio and the face. Fast rap, whispering, heavy effects on the voice and photos with obstructed mouths can look less convincing. The image must be JPEG or PNG and the audio at most 10 minutes. Always make sure you have permission to animate the person in the photo, and avoid using the tool to make someone appear to say or sing things they did not.
Limits and cost for this page, read from the model configuration.
| Model | InfiniteTalk (audio-driven image-to-video) |
|---|---|
| Image | One JPEG or PNG portrait; optional mask image |
| Audio | MP3, WAV or M4A, up to 100 MB and 10 minutes |
| Resolution | 480p or 720p |
| Cost | 480p: 15 credits per 5 s; 720p: 30 credits per 5 s (audio length rounded up) |
| Optional inputs | Prompt (translated to English), mask image, seed |
Ideas for the finished clip.
Give a new single a face-forward video from an artist photo.
Send a birthday or holiday song sung by a portrait, with permission from the person pictured.
Animate a character portrait as the performer of an AI-made song.
Add a singing narrator to a slideshow or a lyric video.
A photo, a track, one video.
Use a JPEG or PNG with a visible face looking toward the camera. Add an optional mask image if you want to limit which area animates.
Front-facing photos with the mouth visible work best
Upload an MP3, WAV or M4A up to 100 MB and 10 minutes. Use a song, a voice-over or a recording from the voice recorder.
Isolate the vocal first for tighter lip sync
Cost scales with the audio length, rounded up to 5-second blocks. Add a prompt to direct the performance.
Test with a short clip at 480p before a full song
Inputs, cost and quality.
Pages for the music that goes with the photo.
あなたに合ったプランを選んでください。隠れた料金や予期せぬ費用はありません。
音楽の旅を始めよう
創造力を解き放とう
プロ品質の音楽制作
