Omnihuman - 1.5 Portrait Animation

Omnihuman

Omnihuman

Omnihuman is ByteDance's audio-driven digital human model. Upload a portrait or source image and a voice track, and it lip-syncs the character to your audio. 1.5 Portrait Animation drives short talking heads (audio under 60s); 2.0 Audio to Video handles longer tracks (4–300s). Billed per second of audio.

1.5 Portrait Animation
2.0 Audio to Video

Configuration

Click to upload audio (under 60s, MP3/WAV/AAC/OGG)

Duration is read from the audio and is the billing basis.

Allowed range 1–59s. Billed per second.

Fast mode trades some fidelity for quicker drafts.

No video generated yet

Upload an image and audio, then click "Generate Video" to start.

🧬 Omnihuman · Audio-Driven Digital Human · Per-second

🧬 Omnihuman 1.5 Portrait Animation

Drive a single portrait with your own audio. Upload one image URL and one audio URL, set the duration (1–59s, audio must be under 60s), and optionally add up to 5 mask images, a prompt (≤300 characters), an output resolution (720 or 1080, default 1080), a fast mode toggle and a seed. Credits bill per second of audio duration.

What is Omnihuman 1.5 Portrait Animation?

Omnihuman 1.5 Portrait Animation (omnihuman-1-5)—an audio-driven talking-head model from ByteDance. It animates a single portrait to lip-sync an uploaded voice track (under 60 seconds), with optional region masks, an optional prompt, 720p/1080p output, a fast mode and a reproducible seed.

🧬 Omnihuman on FuseAITools

Omnihuman is ByteDance's audio-driven digital human family on FuseAITools, offered as two workflows—1.5 Portrait Animation and 2.0 Audio to Video. Both animate a portrait or source image to lip-sync an audio track you upload, and both bill with a straightforward per-second rate multiplied by the audio duration. 1.5 targets short talking heads (under 60s) with region masks, resolution, fast mode and seed controls; 2.0 handles longer tracks (4–300s) from one image with an optional watermark. New users receive 10 free credits on sign-up.

✨ Omnihuman Core Features

Audio-driven lip sync

Animate a portrait or image against your own voice track.

Single-image input

One portrait (1.5) or source image (2.0) drives the whole clip.

Region masks (1.5)

Add up to 5 masks to steer which regions animate.

Wide duration range

1–59s on 1.5, 4–300s on 2.0, billed per second.

Fast mode & seed (1.5)

Draft quickly with fast mode, lock results with a seed.

Cloud on FuseAITools

Per-second credits shown before you generate.

🎯 Built for These Scenarios

Talking-head avatarsVoiceover presentersCharacter dubbingLong-form narrationSocial + explainers

📊 1.5 Portrait vs 2.0 Audio to Video

Dimension1.5 Portrait Animation2.0 Audio to Video
Main inputOne portrait image + one audio URLOne source image + one audio URL
Duration1–59s (audio <60s)4–300s
MasksUp to 5 optional maskUrlsNone
Resolution720 / 1080 (default 1080)Model default
ExtrasFast mode, seedWatermark toggle (default on)
PromptOptional, ≤300 characters
BillingPer second × duration

Duration is read from your audio and is the billing basis for both workflows.

❓ FAQ (Omnihuman)

The 1.5 portrait animation requires audio under 60 seconds—you set a duration of 1–59 seconds, which is also the billing basis. The model lip-syncs the portrait to that voice track.

⚙️ Omnihuman Technical Specs

WorkflowmodelKeyRequiredKey controls
1.5 Portrait Animationomnihuman-1-5One imageUrl + one audioUrl + duration (1–59s)Optional ≤5 maskUrls; prompt ≤300; outputResolution 720/1080; peFastMode; seed
2.0 Audio to Videoomnihuman-2-0One image + one audioUrl + duration (4–300s)Optional prompt ≤300; enableWatermark (default true)

Pricing is per second of the audio duration. Total credits appear on the Generate button.

Omnihuman — Two Workflows

💳 New users get 10 free credits on sign-up. View pricing for subscription discounts and credit top-ups.
🔗 Need a longer voice track? Switch to 2.0 Audio to Video (4–300s) for narration, or stay on 1.5 Portrait Animation for short talking heads under 60s.
🎙️ Driving audio from text first? Generate a voiceover with ElevenLabs Multilingual V2, then upload it as the Omnihuman audio track.