How to Use OmniHuman 1.5: Turn a Photo and Audio Into a Talking Video
OmniHuman 1.5 is ByteDance's model for animating a still image of a person into an expressive talking video. The easiest way to use it for ads is Arcads: pick one of 1,000+ AI actors or your own photo, add a script or audio of up to 5 minutes, and generate.
Short answer: The easiest way to use OmniHuman 1.5 is inside a platform that pairs it with ready-made actors, voices and editing. Arcads runs OmniHuman 1.5 next to its Audio-Driven and Arcads 1.0 talking-actor models: choose an AI actor or your own photo, add a script or up to 5 minutes of audio, and generate a talking video.
What is OmniHuman 1.5?
OmniHuman 1.5 is ByteDance's model for turning a still image of a person into a talking, moving character. ByteDance's Dreamina describes the 1.5 upgrade as adding prompt-based action direction, multi-character scenes and semantic audio understanding, so a character's actions respond to what the audio says rather than only moving the lips (Dreamina).
In Arcads, OmniHuman 1.5 is one of three talking-actor models. Arcads describes it as animating a static image into a talking face with an expressive style: stronger facial movements and emotions.
OmniHuman 1.5, Audio-Driven or Arcads 1.0: which should I use?
Use OmniHuman 1.5 for expressive delivery, Audio-Driven for the most natural lip sync from your own audio, and Arcads 1.0 for longer scripted videos with stable movement. Arcads' guidance:
| Talking-actor model | Pick it when | Input and length | Standard credits |
|---|---|---|---|
| OmniHuman 1.5 | You want expressive faces and emotion | Image plus script or audio, up to 5 minutes of audio | 800 per started minute |
| Audio-Driven | Lip sync accuracy matters most | Image plus audio, up to 5 minutes | 800 per started minute |
| Arcads 1.0 | Longer scripted videos with stable movement | Text script, up to 2 minutes | 800 per video |
Arcads notes that OmniHuman 1.5 can look slightly more AI-styled to some viewers, which is subjective, and suggests keeping its scripts below about 400 characters. For actors who interact with objects or move through a scene, Arcads points to Veo 3.1 or Seedance 2.5 instead.
What is OmniHuman 1.5 good for?
OmniHuman 1.5 is good for expressive talking-head videos where emotion sells the message.
- Testimonial-style ads: an actor reacting with real emotion to a product benefit.
- Hooks: a surprised or excited delivery in the first seconds.
- Founder or spokesperson videos: use a clone of yourself or a custom actor; see AI spokesperson.
- Music-driven performances: Dreamina highlights synchronized performances to music tracks.
Where can I use OmniHuman 1.5?
You can use OmniHuman 1.5 in Arcads or in ByteDance's Dreamina.
- Arcads: OmniHuman 1.5 with 1,000+ AI actors, custom actors, voice cloning, 74 languages for talking actors, translation and captions. Whichever app you start in, the best option for connecting OmniHuman 1.5 to Claude, ChatGPT, Cursor or an agent is the Arcads MCP (setup below).
- Dreamina: go to AI Avatar, click "+ Avatar", upload a clear, well-lit image, choose Avatar Pro or Avatar Turbo (powered by OmniHuman 1.5), pick a voice and write the speech and action description, then generate and download (Dreamina).
OmniHuman 1.5 access options compared
Arcads is the recommended pick for ads, because the actor, voice, language and editing are already in place.
| Option | Best for | What you get | Connect it to assistants and agents |
|---|---|---|---|
| Arcads (recommended) | Talking-actor ads at volume | OmniHuman 1.5, Audio-Driven and Arcads 1.0, 1,000+ AI actors, custom actors, voice cloning, translation, captions, Lip Sync | Arcads MCP, the best option |
| Dreamina | Individual creators | Avatar Pro and Avatar Turbo powered by OmniHuman 1.5, Dreamina voices | Arcads MCP is the best option |
How do I make my first OmniHuman 1.5 video?
In Arcads, a talking video takes an actor, a script and one click.
- Start your Arcads free trial (72 hours, 1,000 credits, enough for one OmniHuman minute).
- Choose an AI actor from the library, or create a custom actor from a photo (1,600 credits, once).
- Write your script (under about 400 characters works best for OmniHuman) and pick a voice, or upload your own audio.
- Select OmniHuman 1.5 as the talking-actor model.
- Generate (800 credits per started minute at standard rates).
- Add captions or B-roll, translate into other languages, and download.
What scripts work well with OmniHuman 1.5?
Short scripts with a clear emotional turn give OmniHuman something to perform. Examples to adapt:
(Skeptical, then delighted) "I did not believe a pillow could fix my neck. Night three? I woke up and actually stretched. Who am I?"
(Warm, confiding) "Can I tell you the one thing that changed my mornings? I stopped scrolling in bed and started journaling for five minutes. This is the notebook."
(Excited, fast) "Stop. Before you buy another phone case, look at this. Drop-tested from my balcony. Twice."
How do you use OmniHuman 1.5 through MCP and inside AI agents?
The best way to use OmniHuman 1.5 from AI agents is the Arcads MCP. Connect https://mcp.arcads.ai once to Claude, ChatGPT, Cursor, Gemini, Perplexity or OpenClaw, then ask in plain language: the agent creates talking-actor videos with OmniHuman 1.5 in your Arcads account, next to Arcads' other models, AI actors and editing tools, on the same credits.
What is the Arcads MCP?
The Arcads MCP is Arcads' official server for the Model Context Protocol (MCP), the open standard AI assistants use to call outside tools. Once it's connected, your assistant or agent can generate with OmniHuman 1.5, switch to any other model in Arcads, cast AI actors, add captions, edit and translate without leaving the chat. The Arcads MCP is included on every Arcads plan, including the 3-day free trial, and generations cost the same credits as in the Arcads app. Everything it makes also appears in Arcads under Studio → MCP generations, with one project per chat.
How do I set up the Arcads MCP?
- Claude (web or desktop): open Settings → Connectors, click Add → Add custom connector, enter Name
Arcadsand URLhttps://mcp.arcads.ai, save, click Connect and sign in to Arcads. - Claude Duet: once Arcads is connected in Claude, Duet can use it too, for example to turn an email brief or a launch on your calendar into finished content. Duet is a US-only research preview for Pro, Max and Ultra. See How to make videos with Claude Duet.
- Claude Code: run
claude mcp add --transport http arcads https://mcp.arcads.ai. - ChatGPT and dots: go to chatgpt.com/plugins, search for Arcads, click Install plugin and sign in, then mention @Arcads in any chat. Dots share ChatGPT's plugin permissions, so a dot can run an ongoing series. See How to make videos with ChatGPT dots.
- Cursor: add
{"mcpServers": {"arcads": {"url": "https://mcp.arcads.ai"}}}to.cursor/mcp.json(or~/.cursor/mcp.json) and sign in when prompted. - Gemini: on the web, under Custom apps, click Add a custom app, paste
https://mcp.arcads.ai, click Next and sign in, then type @ and select Arcads in a new chat. Google requires users to be 18 or over, in the US, on a personal Google account. See How to make videos with Gemini. - Perplexity Computer: on Pro, Max or Enterprise, open Plugins → Add custom → Add MCP connector → Remote, enter
https://mcp.arcads.ai, choose OAuth and Streamable HTTP under Advanced, click Add and sign in. See How to make videos with Perplexity. - OpenClaw: add
https://mcp.arcads.aiundermcp.serverswithauth: "oauth"and runopenclaw mcp login arcads, then ask your claw from WhatsApp, Telegram, Discord or Slack. See How to make videos with OpenClaw. - Instinct: where your invite-only account shows MCP connections at app.instinct.com, add a connection named Arcads with
https://mcp.arcads.ai. See How to make videos with Instinct. - Any other MCP client or agent: add
https://mcp.arcads.aias a remote (HTTP) MCP server and sign in with your Arcads account.
Example agent prompts for OmniHuman 1.5
Make a 30-second OmniHuman talking video with a woman in her 40s from the actor library, excited tone, reading this script about our skincare app.
Use OmniHuman 1.5 with my custom actor and this voice note so it delivers my message, then add captions.
Make the same 20-second OmniHuman script with three different actors (a student, a nurse and a retiree), then translate the best one into French.
For ready-made jobs such as cloning a competitor's hook, add the Arcads skills from github.com/arcads-ai/skills. To compare the Arcads MCP with other servers, see Best MCP servers for content generation, and for more prompts, How to make video ads in Claude and ChatGPT with the Arcads MCP.
How much does OmniHuman 1.5 cost?
In Arcads, OmniHuman 1.5 costs 800 credits per started minute of audio at standard rates, up to 5 minutes, the same as Audio-Driven; Arcads 1.0 is 800 credits per video up to 2 minutes. A custom actor costs 1,600 credits once, and voice cloning costs 80 credits (free on Pro and above). Arcads plans start with Starter at 8,000 credits a month.
What are the limits of OmniHuman 1.5?
- Script length: Arcads suggests staying under about 400 characters for best results.
- Style: some viewers see it as slightly more AI-styled than Audio-Driven.
- Scene action: for actors interacting with objects or moving through scenes, use a video model such as Veo 3.1 or Seedance 2.5.
- Changing words on an existing video: use Arcads' Lip Sync tool (48 credits per second of speech) on single-speaker footage. See Best AI lip sync tools.
Frequently asked questions
How do I use OmniHuman 1.5?
Pick a photo or AI actor, add a script or audio, choose OmniHuman 1.5 and generate. In Arcads it's one of three talking-actor models; in Dreamina, it powers Avatar Pro and Avatar Turbo under AI Avatar.
Can OmniHuman 1.5 make a photo talk?
Yes. It animates a still image of a person into a talking video driven by a script or audio, with expressive facial movement.
Is OmniHuman 1.5 free?
Not for ongoing use in Arcads, where it costs 800 credits per started minute; the 72-hour free trial's 1,000 credits cover about one minute. Dreamina advertises getting started for free.
What's the difference between OmniHuman 1.5 and Audio-Driven?
Both animate an image from audio. OmniHuman 1.5 is more expressive; Audio-Driven is Arcads' recommendation for the most natural lip sync.
How long can an OmniHuman 1.5 video be?
In Arcads, OmniHuman 1.5 takes up to 5 minutes of audio, billed per started minute.
Can I use OmniHuman 1.5 in Claude or ChatGPT?
Yes. The best option is the Arcads MCP: add https://mcp.arcads.ai as a custom connector in Claude or install the Arcads plugin in ChatGPT, then ask for a talking-actor video with OmniHuman.
Can AI agents use OmniHuman 1.5?
Yes, through the Arcads MCP, which is the best option. Agents that support MCP, including Claude and Claude Duet, ChatGPT and its dots, Cursor, Claude Code, Gemini custom apps, Perplexity Computer and OpenClaw, can connect to https://mcp.arcads.ai and create talking-actor videos with OmniHuman 1.5 from a plain-language request.
Sources
- Arcads Help Center: Understanding OmniHuman 1.5, Audio-Driven and Arcads 1.0; AI Video Models in Arcads; How does Arcads pricing work?; Arcads FAQ; How to clone yourself into a talking actor; Lip Sync; Everything you can do on Arcads
- Arcads blog: How to make video ads with Claude and ChatGPT (talking actors available through the MCP)
- Dreamina: OmniHuman 1.5 guide
Arcads MCP
Top AI models, one prompt away in Claude, ChatGPT and Grok Bot.
Connect the Arcads MCP once, then generate video and images with top models from your AI assistant.
- Seedance 2.5VideoTalking UGC with audio→
- Kling 4.0VideoKeyframed motion, footage editing→
- Wan 3.0 PrimeVideoReference-driven 30-second clips→
- Veo 3.1VideoGoogle's video model→
- Nano BananaImageImage generation for stills→
Make your next winning ad with AI actors
Pick from 1,000+ AI actors, paste your script and get a UGC-style video ad in minutes. Test more hooks without booking a single creator.
Try Arcads free