zhenzhen-heygen-avatar5-fal

If the avatar 4 i2v node is "make my photo talk," this one is "pick a professional presenter and press record." HeyGen avatar 5 ships with a catalog of 498 stock avatars - "Abigail Office Front," "Amelia Business Training," that whole LinkedIn-meets-newsroom cast - plus 100+ voices, and this node turns the whole thing into a single ComfyUI input. You pick an avatar, you pick a voice, you type a script, and a polished presenter video comes back.

That's genuinely useful for a specific job: training videos, product explainers, internal comms, anything where you want a human presenter but don't want to book one. It's also the most "turnkey" node in the Comfyui-zhenzhen pack - no image input at all, because the avatar already exists. Like every node here it's a closed-model API call through Zhenzhen's fal proxy: per-clip credits, no local weights, key in the widget.

How it works

The node submits the script (prompt) plus your avatar/voice choices to HeyGen's avatar 5 endpoint via the fal proxy, polls until done, and returns a VIDEO plus video_url/response. Because the avatar is predefined, there's no face upload - the two big inputs are:

• avatar - the giant dropdown (498 choices, server_default lets fal choose). If you know the exact avatar ID from the HeyGen catalog, custom_avatar overrides the dropdown.
• voice - 100+ voices, server_default again for fal's pick, custom_voice for a specific ID. If audio_url is provided, the voice is ignored and the avatar lip-syncs to that audio instead.

Inputs that matter

• prompt - the script. This is what the avatar says; quality of delivery follows quality of writing.
• avatar / voice / custom_avatar / custom_voice - who's talking and with whose voice.
• fit - cover (fill the frame, may crop) vs contain (whole avatar visible, may letterbox).
• remove_background - cut the presenter out; handy for compositing over your own background later.
• caption - subtitles on/off. output_format - mp4 or webm. resolution - 720p / 1080p / 4k. aspect_ratio - 16:9 / 9:16 / 4:5 / 5:4 / 1:1 / auto.

Installing it

ComfyUI Manager → search Comfyui-zhenzhen → install, or:

cd ComfyUI/custom_nodes
git clone https://github.com/T8mars/Comfyui-zhenzhen

Restart, paste your api_key into the widget. Nothing to download - the whole job happens server-side.

Common issues

• Avatar not matching the vibe - the 498-name dropdown is hard to scan. Use custom_avatar if you have a catalog ID; otherwise server_default is a decent neutral pick.
• 4k being slow/pricey - yes. This is where fal's pre-deducted billing is most visible; iterate at 720p, export at 4k.
• Voice ignored - you supplied audio_url; that's by design. Remove it to get TTS with the selected voice.
• The pack standard gotchas - empty key, 443/VPN-in-TUN-mode for the overseas API, vendor 500s needing a re-run, and seed being cache-control only.

One thing to keep straight: this node is a stock presenter tool, not a face-clone. If you need a specific real person talking, that's the avatar 4 i2v node's job (with all the moderation caveats that implies); if you need "a believable human says the words," this is the one.