ByteDance

Mode: video · Models: 2

Vendor: BytePlus · Official API docs: OmniHuman 1.5 overview

ByteDance models on the BytePlus Vision AI platform. OmniHuman 1.5 is an audio-driven avatar model — give it a single portrait image plus an audio clip and it generates a talking/performing video (expression and motion are driven by the audio, not a text prompt). A separate ByteDance Video Enhance denoises, colour-corrects and super-resolves existing footage up to 8K, and can convert frame rate.

Models

id	Name	Input type
bytedance-omnihuman-v1.5	ByteDance OmniHuman	i2v
bytedance-video-enhance	ByteDance Video Enhance	v2v

CLI

# audio-driven talking avatar: portrait image + audio clip
gen-ai generate -m bytedance-omnihuman-v1.5 \
  -i ./portrait.jpg -a ./speech.mp3 \
  -p "subtle head movement, slow camera push-in"

# restore and super-resolve an existing clip to 4K at 60fps
gen-ai generate -m bytedance-video-enhance --video ./clip.mp4 \
  -r 4k --fps 60 --quality professional

MCP

{ "name": "picsart_generate",
  "arguments": {
    "model": "bytedance-omnihuman-v1.5",
    "imageUrls": ["https://example.com/portrait.jpg"],
    "audioUrl": "https://example.com/speech.mp3",
    "prompt": "subtle head movement, slow camera push-in"
  } }

{ "name": "picsart_generate",
  "arguments": {
    "model": "bytedance-video-enhance",
    "videoUrl": "https://example.com/clip.mp4",
    "resolution": "4k",
    "fps": "60",
    "quality": "professional"
  } }

Parameters

Full parameter surface for every model, sourced from gen-ai models info <id> --json. CLI flags show the primary short form; the canonical --kebab-case long form always works too.

bytedance-omnihuman-v1.5 — ByteDance OmniHuman

Try bytedance-omnihuman-v1.5 in Playground ↗

Input type: i2v

Param	CLI flag	Type	Values
prompt	-p	text	free text
imageUrls	-i	file	required image (up to 1)
audioUrl	-a	file	required audio

Notes: OmniHuman 1.5 derives emotion and lip-sync from the audio, so prompt is optional and only steers camera/motion. Video Enhance needs only a source video; every other param refines the restore.

bytedance-video-enhance — ByteDance Video Enhance

Try bytedance-video-enhance in Playground ↗

Input type: v2v

Param	CLI flag	Type	Values
videoUrl	--video	file	required video
quality	--quality	enum	standard · professional (default standard)
resolution	-r	enum	source · 720p · 1080p · 2k · 4k · 8k (default source)
fps	--fps	enum	30 · 60 · 120 (default 30)
scene	--scene	enum	common · ugc · short_series · aigc · old_film (default common)
bitrateLevel	--bitrate-level	enum	low · medium · high (default medium)

Pricing

gen-ai pricing bytedance-omnihuman-v1.5

Cost scales with the duration of the generated video (driven by the length of the input audio clip).