Supports MP3, WAV, AAC, OGG, WEBM, FLAC, or M4A up to 200 MB via the secure proxy.
Upload audio to estimate credit usage.
Use a clear front-facing portrait in JPG, PNG, or WebP up to 40 MB.
Limit video generation to a subject by providing a binary mask. Leave blank for automatic detection.
Prompts can shape posture, lighting, or energy. Leave blank for natural delivery.
480p is ideal for drafts. Use 720p when you are ready to publish.
Use the same seed to match expressions across takes. -1 randomizes each run.
Why teams trust InfiniteTalk
The pipeline keeps face integrity while following your performance cues.
Phoneme-accurate lip sync
Aligns each syllable with the voice track, preserving timing and breathing pauses.
Expressive performances
Captures subtle eye lines, nods, and torso sway so avatars feel engaged rather than robotic.
Identity locked every frame
Maintains hairstyle, wardrobe, and lighting across long takes—even during profile turns.
Up to 10-minute renders
Generate long-form explainers or product demos without stitching multiple clips together.
How to create an InfiniteTalk video
Follow this sequence to keep processing predictable and high quality.
- Step 1
Upload the polished voice track
Use a clean mono mix with minimal reverb. InfiniteTalk analyses the waveform to predict phonemes.
- Step 2
Choose a clear portrait
Frontal or three-quarter images work best. Upload an optional mask if multiple people appear.
- Step 3
Set resolution and guidance
Pick 480p for drafts or 720p for final delivery. Add a prompt for posture, mood, or camera framing.
- Step 4
Submit and let InfiniteTalk animate
We bill in 5-second increments. You receive a status link plus download-ready MP4 once complete.
Production notes
Maximum clip length is 10 minutes. Uploads longer than 600 seconds are rejected before billing.
For group photos, upload a binary mask so InfiniteTalk knows which subject should articulate.
Prompts accept creative direction like “gentle smile”, “studio key light”, or “subtle nodding pacing”.
Credits are calculated in 5-second blocks. 720p renders cost roughly 2× more than 480p previews.