OmniHuman 1.5: The Lip-Sync Engine Behind Talking Actors

OmniHuman 1.5 is ByteDance's audio-driven human animation model: give it a single still image of a person and a voice track, and it generates a video of that person speaking the lines — mouth, head movement, and natural gestures included. It is the engine behind Flovaly's Talking Actors mode, where a character, a script, and a voice become a lip-synced clip with no filming involved.

What is OmniHuman 1.5?

OmniHuman is a family of human animation models from ByteDance, and 1.5 is the version Flovaly runs for lip-sync. Unlike most lip-sync tools, which need an existing video to modify, OmniHuman is designed to work from a single still image plus an audio track. The model listens to the speech and animates the whole figure to match — not just the mouth, but head tilts, blinks, and the small hand and shoulder movements that make a delivery read as human rather than puppet-like.

That mechanism matters. Because no base video is required, there is nothing for the animation to fight against: the character's appearance comes entirely from the still, and the performance comes entirely from the audio. In our experience it produces the most convincing mouth-to-voice alignment of the lip-sync options we have run — which is why it is Flovaly's default for the job.

How Flovaly uses it in Talking Actors

When you run a Talking Actors generation, Flovaly does not generate a base video first. The pipeline is leaner than that:

• Your script is converted to speech in the voice you picked.
• The audio is cleaned before animation — silence trimmed, loudness normalised — so the model gets a tidy performance to work from.
• OmniHuman 1.5 animates your character's reference image directly from that audio, producing the finished talking clip.
• Captions are burned in by default (they lift performance on short-form social), and you can switch them off per generation.

Any actor works as the input — a premade character from the library or a custom one you have built. If you have not created an actor yet, start with Create your own custom actor.

How to make a talking-actor clip

1. Open the Studio — it opens in Talking Actors mode by default.
2. Click Add actors in the composer footer and pick a character. Faces that are clear and well lit in the reference image animate best.
3. Write the script in the main textarea. This is the literal line the actor speaks, and punctuation drives pacing — contract your words and break long thoughts into two sentences. A character counter shows the limit.
4. Pick the voice in the audio settings panel (the gear icon); the inline voice row beneath the composer always shows the current selection and accent.
5. Check the live credit cost next to the Run button, then run. The clip lands in your gallery when it finishes.

For the full control-by-control walkthrough — including tips for a natural delivery — see the Talking Actors guide.

Pricing in credits

A talking-actor clip starts at 7 credits: 5 for a five-second generation at 720p plus 2 for the voiceover. Ten and fifteen-second durations cost more, and 1080p output applies a 1.5× multiplier. As everywhere in the Studio, the exact figure for your current settings is shown next to the Run button before you commit, and failed jobs are refunded automatically. For how credits map to plans, see Credits, plans, and billing. For how that 7-credit figure sits alongside every other generation type on Flovaly, see our full AI video ad cost breakdown.

OmniHuman vs Re-voice

Flovaly has two lip-sync pipelines, and they solve different problems. OmniHuman generates a brand-new talking video from a still image and audio — use it when you are starting from a character. Re-voice, powered by sync-lipsync v2, takes an existing video and replaces its audio and mouth movements with a new script — use it when you already have footage you want to keep. We cover the second pipeline in Re-voice AI video with sync-lipsync v2.

When to choose Talking Actors

Reach for Talking Actors whenever the spoken script is the point: a UGC ad, a hook, a founder testimonial, a product announcement. The actor holds frame and the camera stays steady, so the voice and face carry the clip. If you need camera movement, action, or an environment doing the talking, generate with a base video model instead — see our 2026 AI video model comparison for which one fits the brief. For tips on picking a voice and writing a script that reads as spoken rather than typed, see AI UGC ads that don't look AI. If you are weighing Talking Actors against a dedicated avatar platform, see Flovaly vs HeyGen for an honest per-minute cost and feature comparison.

FAQ