Higgsfield Standard is the presenter model on this canvas. Its distinguishing feature is identity lock from reference images, which means the same figure can appear across many clips without the drift that breaks character consistency on general video models.
One naming note: the model is Higgsfield Standard, not "Soul 2.0". That spelling appeared across a lot of secondhand coverage including, until recently, parts of this blog.
TL;DR
- 95 credits for 5 seconds at 720p, 189 at 1080p
- Editing modes run 189 and 378
- Identity lock from reference images is the reason to use it
- Consent: a presenter built on a real person needs their documented permission for the specific use
- For a founder or a real spokesperson, a phone camera wins. This is for volume, not for trust
What it costs
Canon 8frame prices, a credit is $0.01 at pack rate:
| Mode | Spec | Credits |
|---|---|---|
| Higgsfield Standard | 5s at 720p | 95 |
| Higgsfield Standard | 5s at 1080p | 189 |
| Edit | per pass | 189 |
| Edit, heavier | per pass | 378 |
At 19 credits per second at 720p it sits mid-tier. The comparison that matters is not price but capability: nothing cheaper on this canvas holds a presenter identity across clips.
What identity lock actually solves
The characteristic failure of character work on general video models is drift: the same described person comes out differently in every generation, and by the fourth clip a viewer has lost track of who they are watching.
Identity lock anchors the appearance to reference images, so a presenter figure stays recognisably one figure across a series. That is the difference between a set of clips and a character.
What it still will not do: hold perfectly at every angle. Profile and back views drift more than front and three-quarter. Build sequences from the angles that work rather than fighting for full coverage.
The consent rule
If the presenter is built on a real person's likeness, that requires the person's documented permission for the specific use. Someone who agreed to reference images for an internal training module did not agree to a product advertisement.
Never a public figure. Never a photograph you found. Never a colleague who has not signed something.
Serious platforms enforce this with recorded verification where the subject states consent aloud, and that step is what makes the capability defensible rather than a problem waiting to happen.
The clean alternative: generate a presenter who is nobody. A figure that resembles no real person carries no consent question at all.
Where it fits
Volume scripted content. Internal training, onboarding, documentation video, product explainers where a presenter adds structure and the same figure should recur.
Multilingual versions of the same script, with a native speaker checking every language before it ships.
A recurring brand figure for high-volume content where filming every variant is impractical.
Where it does not
Anything where trust is the product. Testimonials, founder stories, case studies, recruitment. A synthetic presenter in a credibility document undermines the exact thing the video exists to build, and viewers increasingly spot it.
A founder video. Film yourself. A phone, a window, two minutes. It outperforms any avatar, costs nothing, and carries no consent question.
Lip sync on existing footage of a real person. That is OmniHuman's job, and it is the better route for localising your own recorded material.
The workflow
- Prepare reference images. Consistent lighting, a few angles, clean background.
- Generate a test clip at 720p, 95 credits, and check identity holds.
- Write the script to a word count. 150 words per minute of finished video.
- Generate the voice separately if you want control: ElevenLabs TTS at 13 credits per 1,000 characters.
- Keep framing consistent across the series, since that is what makes identity read.
- Generate at delivery resolution. 1080p at 189 if it needs it, because there is no video upscaler on this canvas.
A five-clip training sequence at 720p with narration: 475 credits plus about 40 for voice, roughly $5.15.
FAQ
How much does Higgsfield Standard cost? 95 credits for 5 seconds at 720p, 189 at 1080p.
Is it Higgsfield Standard or Soul 2.0? Higgsfield Standard. The other name is a widespread error.
Do I need consent for a presenter? If it is built on a real person's likeness, yes, documented and specific. A generated figure who is nobody avoids the question.
Should a founder use an avatar? No. Film yourself; it is free and it is the thing that converts.
Lock the identity for volume, film the person for trust. The 8frame canvas is free and unlimited, and generation is paid from $19/month.