I Auditioned 2 AI Hosts for My Faceless Channel. Neither Exists (Higgsfield AI Influencer)

0:00 These two want to host this channel. Neither of them exists. This is Omar. This is Amal. And both of them are auditioning with the same walking reference. This channel has never had a

0:09 host on camera, and I don't own a camera. So, let's hold an audition anyway. Two characters built from scratch with a screen test for each. Then the comments get the casting vote.

0:19 We'll go through the character choices, the motion, and the outfit and location changes. Every recorded price goes into a receipt at the end. And when something changes that should have stayed the

0:28 same, we'll put the two versions next to each other. This video is sponsored by Higgsfield. These are Higgsfield's examples of the character workflow. Our audition starts with the same studio,

0:38 but with two candidates built for a developer channel. Here is AI Influencer Studio. The builder offers average, bold, and extreme character types. That gives the audition a fairly wide range.

0:49 The extreme candidates didn't make the short list. It's hard to host a debugging tutorial when your hair is the main event. The useful part is the control underneath. Gender, build, hair,

0:59 clothing style, age, face details, and accessories are separate choices. You can see the selected tiles as the list scrolls. That makes it easier to decide what should identify a character before

1:10 making the video. There is an optional photo upload at the top. This audition uses the builder's choices instead. No performance was filmed for it, either. The walking reference comes from

1:20 Higgsfield's own preset gallery. Start with Omar, the on-call SRE. His final sheet shows an adult male character, a slim build, a denim shirt, and headphones. The headphones are the

1:31 clearest signature. They give him something recognizable when the shot gets wider. The recording shows those choices being made, rather than just jumping to a finished portrait. Male,

1:41 build, hairstyle, then the color and style controls. Later, thick brows, stubble, and headphones are selected. There is a face close-up and a full body view in the result. One small detail

1:52 worth checking. The pointer hovers over heavy eye bags, but that tile isn't visibly selected. So, that doesn't count as a confirmed setting. A A hover is not a click. The finished face also looks

2:02 less exhausted than an on-call casting description might suggest. The character sheet's generate button shows 1.12 credits. That's the number for this recorded setup. Keep the portrait and

2:13 the full-body view together. They give the next step both an identity and an outfit to work from. Now, Emma, the staff engineer. Her strongest visual difference is the braids. The final

2:23 controls show an athletic build, a mature age setting and dark skin. Her sheet gives her a striped top, dark jacket, and glasses resting on top of her head. You can see the revisions to

2:33 the choices while the builder scrolls. Then, glasses and headphones are selected. Check the result rather than assuming every accessory landed exactly where expected. Here, the glasses sit on

2:43 her head and the hair and headphones obscure the ear details. Emma's sheet also shows 1.12 credits on generate. Both completed sheets now sit in the same history. That is our casting pair.

2:54 Save those identities because the next stage should use these characters instead of designing two new faces for every clip. For the audition, switch to motion and choose motion transfer. This

3:04 is Genjutsu. There are two separate ingredients on the left, the performance reference and the character image. One supplies the take, the other tells the tool who should appear in it. In

3:14 Higgsfield presets, the selected reference is the walking tile beside the pool. Its name isn't readable here, so let's call it the pool walk. The reference already contains several

3:24 camera setups. This is not a single uninterrupted shot. Add Omar's save sheet, check the prompt, and leave the recorded quality at 1080p. The button shows a discounted price of 132 credits.

3:36 The older price is crossed out. Then, generate is clicked and the new take appears in history. Here is the downloaded Omar take. The face, headphones, and clothing stay

3:45 recognizable as the view changes. His hands mostly stay in his pockets, so So particular clip doesn't tell us much about complicated finger movement. The camera sequence is easy to compare. For

3:56 Emma, keep that same performance reference and change the character image. Her prompt is updated, the quality is still 1080p, and generate again shows 132 credits. The actual

4:07 click and the completed take are both in the recording. Now put them beside each other starting at the same moment. The opening side view and the next front view line up closely. There is the wide

4:18 pool shot, then the profile. The final cut is one source frame apart. Close, but not perfectly identical. For choosing a host, the face is only part of it. Watch whether that identity

4:29 survives a wider view, a turn, and a close-up. Omar's headphones are easy to follow. Emma's braids stand out. Exact step shapes differ even with the same source performance. The screen test

4:40 starts with clothing. Switch to object swap, select Omar's original walk, and add this navy suit reference. The prompt asks for the clothes to change while the person, movement, and background stay.

4:52 The recorded button price is 143 credits. Play the original and the suit version from zero together. The shirt and rolled trousers become a suit and white shirt. Omar still has his face and

5:03 headphones, and the pool sequence broadly remains. There are tiny cut differences, so check transitions as well as the still frames. Emma gets the same clothing test with a cream suit

5:13 reference. Select her original take, add the outfit image, and submit the change. This one also shows 143 credits. The result gives her a cream jacket and trousers with a dark top. The braids and

5:25 headphones are still recognizable. This is the strongest kind of comparison in this session. The broad scene and walking sequence stay familiar while the wardrobe changes. It also gives a useful

5:36 check on the face when the original clothing stops doing the recognition work. Next, send each original walk to the same blue server aisle location. These start from the original outfits,

5:46 not from the suit swaps. The location reference is visible in the recorded picker. Both generate buttons show 143 credits, and both results move into the server setting, but the camera changes,

5:57 too. Omar's server version starts with a different framing. Omar's version moves into a much tighter face shot, then opens into the aisle. The new location worked here. Exact camera preservation

6:08 did not. That matters if your edit depends on matching every cut. There is also a product test. This time, the source is suited Omar. Add the plain black mug reference, ask for it in his

6:18 free hand, and submit the 143 credit change. Compare against the suit version, and the mug is the new element you can actually isolate. Finally, reuse Omar's save sheet on a second motion.

6:30 The recording selects that existing character, checks a 720p output, and shows 70 credits before generate. In the result, he handles sunglasses outside a brick building. Same saved identity,

6:42 different performance. Check those smaller hand details closely. Now, the receipt. Before the session, the balance is 13,963.09. After, it is 12,884.73.

6:57 Subtracting the second number from the first gives a balance drop of 1,078.36 credits. The prices recorded on the buttons are separate evidence. Two character sheets at 1.12,

7:09 two walks at 132, five swaps at 143, the second motion at 70. Those displayed prices add up to 1,051.24. That leaves 27.12 credits beyond the recorded price list.

7:25 The footage doesn't itemize that remainder, so it would be wrong to assign it to a specific job. No second try is visible in these recordings. The balance drop and the button price total

7:34 are different measurements. For a next attempt, choose one person with the full body visible and a a continuous movement. That reduces what needs checking. Give the character a clear

7:44 signature, then inspect the generated result at wide and close distances. When changing a location, check the camera separately. A convincing face does not guarantee a matching take. The verdict

7:55 from this audition, the builder makes two distinct reusable characters and the outfit comparisons hold the broad scene well. The location changes are usable as new scenes, but they change the framing.

8:07 These are short visual tests. There is no speaking host test here, so they don't prove either candidate can carry a full narrated episode. Now the casting vote. One for Omar, two for Uma.

8:18 Or zero to keep this channel faceless. Neither of them exists, but the choice can stay real. To try the same workflow, use the Higgsfield link in the description and pinned comment. Which

8:29 one gets the job?