Higgsfield AI Influencer's 19 settings vs a Sume avatar and face swap

Higgsfield's updated AI Influencer tool builds a character from 19 settings and then puts that character into an existing video. Sume has no 19-setting builder. A Sume avatar starts from a text prompt, a reference photo, or three structured traits (ethnicity, sex, age), costs a flat $0.95, and then either speaks a script you write or has its face applied to a short source clip with Face swap (Beta).

The rest of this page maps each Higgsfield step to the closest Sume call, so you can decide which tool fits a recurring character.

What did Higgsfield actually launch?

Higgsfield's blog describes a Character Builder with 19 settings, including character type, gender, ethnicity, age, skin color, build, proportions, features, hairstyle and style. To get the character into footage, it hands the character to Genjutsu, which offers Motion transfer and Objects swap on a reference video of 4 to 30 seconds at 480p, 720p or 1080p. Semafor reported the update on October 7, 2026 and noted that the characters are deliberately stylized rather than photoreal.

How do I create a reusable character on Sume?

Send one request to the Avatar 1.0 route with an avatar_handle and an input. The input is one of three types. prompt is free text. photo takes a fetchable public HTTPS image URL. props takes exactly three fields: ethnicity (Asian, South Asian, Southeast Asian, Black, Hispanic, Middle Eastern, White or Wasian), sex (male or female) and age (20 to 80).

Handles are 2 to 30 characters, stored without the @, and cannot start with the reserved sume_ prefix. Creation is a job: poll /v1/jobs/{id}/status, then read the result. The finished handle is what every later call uses, so the same face comes back each time you reuse it.

Everything beyond the three props fields (hair, proportions, accessories, an exaggerated art style) has to go through the prompt or a reference photo. The docs do not promise a particular look from a prompt, so generate, review the result, and keep the handle only when you like it.

curl -X POST https://api.sume.com/v1/avatar-1.0/generate \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: character-001" \
  -d '{
    "avatar_handle": "mascot_host",
    "input": { "type": "props", "ethnicity": "Asian", "sex": "female", "age": 28 }
  }'

How do I put that character into an existing video?

Use Face swap (Beta): POST /v1/models/sume/avatar-face-swap/v1.0/runs with avatar_handle, a public HTTPS video_url and a required quality of standard, plus or max. It applies the avatar's face to the source clip. The docs plan for source videos of roughly 4 to 15 seconds with usable audio, and the endpoint takes no prompt, transcript, duration or aspect-ratio field.

That is narrower than Genjutsu. Face swap changes the face; it is not a Motion transfer or Objects swap, and it will not hand a character a reference clip's choreography. If you need a 30-second reference, cut the part you want first with video trim at $0.02 a job.

What does each route cost on Sume?

Rates below come from Sume's rate card. Face swap reserves the 15-second maximum at the Avatar Video no-product rate, so a short clip is held at the full figure while the job runs.

When should the character talk instead of being swapped in?

If the goal is a recurring presenter rather than a remake of someone else's clip, skip the swap. Generate avatar video takes a script or ordered video_inputs that total 4 to 60 seconds, in 1:1, 3:4, 9:16, 4:3 or 16:9, and returns a 720p MP4. A 10-second clip at standard is about $1.84.

Use Higgsfield-style remaking when the movement of a reference clip is the point. Use a Sume avatar when the same named character needs to say new lines week after week and you want the whole thing behind one API key.