If you are about to release a single and you do not yet have a stable on screen identity that carries across every video, this is the setup you do once and reuse forever.
An AI music video character in Echonos is a saved on-screen identity built from up to four reference photos and a name. It lives in your Vault and gets reapplied to every video you generate, so the same face and silhouette appear across every release instead of drifting from track to track.
An AI artist persona in Echonos Characters is a saved on screen identity built from up to four reference photos and a name. It lives in your Vault and gets reapplied to every video you generate, so the same face and silhouette appear across every release instead of drifting from track to track.
What is an AI music video character (Echonos Characters explained)?
An artist persona is the human or styled character at the center of every music video you release, treated as a persistent asset rather than a fresh prompt each time. In Echonos, the persona is stored as a Character document and attached to a generation through the selectedCharacter field. The same record powers every video the persona appears in.
This matters because most general purpose AI video tools treat each generation as an independent draw. You upload a song, write a prompt, get a video, and the artist on screen looks like one specific person. You upload the next song, write a similar prompt, and the artist looks like somebody else. There is no memory between generations. A persona, in the Echonos sense, is the memory.
Why an Artist Persona Is Different From a Generic AI Avatar
A generic AI avatar is a one off character generated for a single render. You write a description, the model produces a person, and that person exists for the length of the clip. The next clip starts from zero.
A persona is different in three ways. It is saved. It is named. It is reused. Once you have a persona in Echonos, you select it from the Characters surface during creation and it gets applied through the pipeline as part of the brief. The result is that the figure on screen in your second video is the same figure as in your first, not a similar one.
This is the difference between a character who appears in one frame and an artist who appears in a catalog. Avatars are disposable. Personas compound recognition.
How a Persona Anchors Your Visual Identity Across Every Release
When a listener sees your fifth release, the cheapest signal you can give them is "this is the same artist they liked on release one." Visual recognition is faster than name recognition. A familiar face triggers a micro pause that often determines whether someone hits play or scrolls past.
The persona is the mechanism that keeps that recognition signal stable across a release schedule. The face does not have to be biometrically identical in every frame. It has to read as the same person to a casual viewer scrolling Spotify Canvas, YouTube, or TikTok with the sound off. The persona makes that work without you re briefing the look every release.
Why indie artists skip character setup (and what that costs)
The persona setup takes about ten minutes. It is the single highest leverage ten minutes you will spend on your visual rollout, and most indie artists skip it because it does not feel like creative work. It feels like data entry.
The cost shows up later. You release the first single, you write a long descriptive prompt to lock in the look, the result is decent, and you ship it. Three weeks later you release the next single. You write a similar prompt from memory, but the seed is different, the wording is slightly off, and the artist on screen is not quite the same person. By release four the catalog reads as four different acts, and the recognition equity you have been paying for with every release has gone to zero.
A saved persona prevents that drift entirely. It is the difference between rebuilding the same character in your head every time and pointing at the same record in your Vault every time.
How Inconsistent On Screen Identity Slows Streaming Discovery
Streaming discovery rewards repeat engagement signals. Listeners who watch your Canvas loop, click into your artist page, and stay through the song train the algorithm that your catalog is worth surfacing. A consistent on screen identity multiplies that effect because returning listeners process new releases as familiar before they have read a caption.
Inconsistent identity has the opposite effect. Returning listeners do not recognize you, the micro pause does not happen, watch through drops, and the algorithm reads that as soft demand. You can write the strongest single of your career and it will under perform if the visual on top of it does not connect to your previous releases.
For a deeper read on why this is a brand problem and not just a visual one, see our pillar on consistent character ai (pillar).
What You Need Before You Open the Echonos Characters Builder
Before you click into Characters and tap Add Character, gather four things. The setup goes faster and the result is better when these are ready in advance.
First, four reference photos. The Characters builder offers four slots: a required Headshot, plus optional Full Body, Left Profile, and Right Profile. The Full Body slot is sized for 9:16, which matches the only aspect ratio Echonos currently ships, so a vertical full body shot is ideal. Common image formats (PNG, JPG, WebP, HEIC and several others) are accepted, up to 10 MB per image.
Second, a persona name. The field accepts up to 100 characters. Use a name you will recognize at a glance from a Vault grid. If your stage name is one persona and an alternate styled version is another, name them in a way that disambiguates ("Mara Studio" versus "Mara Live," for example).
Third, a clear visual reference for the look you want locked. Hair, signature wardrobe, characteristic styling. The reference photos do most of the work, but knowing in advance which traits should never change makes it easier to choose photos that show those traits cleanly.
Fourth, account credits if you intend to test the persona on a real generation. New accounts get 250 free signup credits, and Echonos charges flat fees per operation (a full Engine generation is 200 credits regardless of song length). A signup balance covers one full test generation with a little room left over for a Studio scene fix.
Reference Photos, Style Notes, and Mood, What Actually Helps the AI
Not every photo is a good reference. The pipeline is reading these photos to anchor face, body, and silhouette across generations, so the shots that work best are the ones that show those clearly.
Use clean, well lit photos with the subject centered and the face unobscured. Avoid heavy filters, group photos, photos where the artist is partially turned away, or photos where shadows hide one side of the face. The headshot is the most important single image because it carries the face that the model will reproduce in close ups across every video.
For the optional slots, the goal is variety. The Full Body slot tells the pipeline what the silhouette and proportions look like. The two profile slots give the model the side angles it needs to render the artist convincingly when the camera is not facing them straight on. You do not need all four. A strong headshot alone produces a usable persona. Adding the others tightens consistency in shots where the camera moves around the subject.
A small thing that matters more than people expect: pick photos that already match the styling you want the persona to wear. If you upload a headshot in a black hoodie, the persona will tend to read as "artist in dark casual wear." If you upload a high contrast studio portrait in a tailored coat, the persona will tend to read as "studio polished." The references are not just identity. They are also wardrobe and mood.
Step by step: Building your first character in Echonos
The full flow is four steps inside the Characters surface. Plan on ten minutes start to finish, less if your reference photos are already on your machine.
Step 1: Open Characters and Tap Add Character
Open Echonos and navigate to the Characters surface. The Characters tab is the default view. The first tile in the grid is Add Character, with a plus icon. Tap it. A modal opens with the title "Add Character."
The modal has three regions. At the top is the Character Name input, with a 100 character counter on the right. In the middle is a Reference Images grid. At the bottom are the Cancel and Create Character buttons.
If you ever want to revise an existing persona, tap any existing tile in the grid instead of Add Character. The same modal opens with the title "Edit Character" and your previous reference photos already filled in. From there you can swap any slot, rename the persona, or delete it entirely.
Step 2: Upload Your Reference Photos to the Four Slots
The Reference Images grid has four upload targets arranged in a deliberate layout. The largest target on the left is the Full Body slot, sized to a 9:16 frame. To its right, three circular slots stack vertically: Headshot (marked Req), Left Profile, and Right Profile.
Tap the Full Body rectangle and select your vertical full body reference. The image fills the frame. Tap the Headshot circle and select your face shot. Repeat for Left Profile and Right Profile if you have side angles ready.
The interface shows a refresh icon when you hover a filled slot, so you can swap any image without resetting the rest of the form. The footer text under the grid confirms the rules: common image formats (PNG, JPG, WebP, HEIC and several others) up to 10 MB per image, and only the headshot is required.
If only the headshot lands, you still get a working persona. The pipeline will use the single reference for face anchoring across the video. Adding the other three slots gives the model more to work with on body shots, profile turns, and full frame compositions.
Step 3: Name the Persona and Lock the Visual Traits That Should Never Change
Type the persona name in the input at the top of the modal. The 100 character limit gives you room for stage name plus a qualifier ("Echo Hart, Studio Era"), which becomes useful later when one artist runs multiple personas across different release cycles.
The naming step is where you also decide, mentally, which traits the persona is committing to. The reference photos are doing the technical lock. Your job is to know what you uploaded and to remember not to fight it later in the prompt. If the reference shows shoulder length hair and a leather jacket, do not ask the prompt to put the same persona in a buzz cut and a hoodie in your next video. The persona will resist, and the result will look off.
The strongest practice is to think of the persona as an established visual identity, not a starting point. Once it is saved, you describe scenes around it in your prompts, not the persona itself. For more on layering visual locks on top of the persona, see our companion piece on style consistency locks.
Step 4: Tap Create Character and Confirm It Saves to Your Vault
When the headshot is filled and the name is typed, the Create Character button activates. Tap it. The modal shows "Uploading images..." with a spinner, then "Creating character..." Once processing finishes, the modal closes and the new persona appears as a tile in the Characters grid.
Behind the scenes, every reference image you uploaded has been stored in your Echonos Vault, the persona has been registered as a Character document, and a thumbnail has been generated for the grid view. The persona is now available as an asset across every video you create from this account.
You can verify the save by scrolling the Characters grid and looking for the tile with your persona name. Tap it to confirm the references are stored. Hover and you will see a pencil icon, which opens the Edit modal if you ever want to update the persona later.
For more on how the Vault stores and reuses creative assets across releases, see our guide on Echonos Vault asset management.
How to test a character before a real release
Before you trust a new persona on a single you actually care about, test it on a short generation. The cost is small and the information is high.
Open the creation flow, attach a short audio clip you do not mind spending a few credits on, and select your new persona from the Characters picker so it lands in the selectedCharacter field. Write a simple, neutral prompt that gives the pipeline room to render the persona without contradicting it. Something like "the artist performing in a moody warehouse, neon side lighting, slow handheld camera" works because it does not over specify the subject.
Generate the video and watch for three things. Does the face match your reference across multiple shots? Does the silhouette hold steady when the camera moves? Does the persona read as a coherent identity from the first frame to the last? If all three hold, the persona is ready for a real release.
If something drifts, the fix is almost always in the references, not the prompt. Common patterns: the headshot was too soft and the model pulled in details that were not there, the Full Body slot was missing so wide shots invented a new silhouette, or the wardrobe in the references was too varied so the persona never settled on a look. Open the persona in Edit mode, swap the weak references, and run the test again.
Once a persona passes the test, leave it alone. The whole point of the Vault is that you do not rebuild the look every release. From this point on, your creative work is in the prompt and the music, not in re briefing the artist who appears in the video.
A Quick Note on What Is and Is Not Included Today
Two product details worth knowing before you build your first persona.
Generation is currently 9:16 vertical only. The Full Body slot in the Characters builder is sized to that aspect deliberately. Personas you build today will continue to work as additional aspect ratios ship, because the persona record is the input and the aspect ratio is a separate output setting.
The live subscription tier today is the Pilot Plan at thirty dollars a month with 750 credits, and higher volume tiers for active artists and labels are listed as coming soon. New accounts receive 250 signup credits. Persona creation itself does not consume credits. Credits are spent only on generation as flat fees per operation (a full Engine generation is 200 credits regardless of song length). That means you can build, edit, and refine personas as much as you like before committing credits to a real render.
What to do after your first character is saved
You have a saved persona. The next move is to treat it as the foundation of your visual rollout for at least the next release cycle. Three habits make that work.
First, attach the persona to every video you generate this cycle, even the ones you might not ship. The point is consistency across everything you put out, including teasers, lyric clips, and Canvas loops. Consistency compounds. One off variations dilute it.
Second, leave the persona alone unless you are intentionally pivoting eras. Editing the persona mid cycle resets the visual lock and your audience loses the recognition thread. Save edits for the boundary between release cycles, not the middle of one.
Third, when you do start a new cycle and you want a different visual era, build a new persona instead of editing the old one. Multiple personas are supported. Saving the old persona alongside the new one means you can revisit older releases without the visual identity rewriting itself in your archive.
When you are ready, head to the Characters surface, tap Add Character, and spend the ten minutes. The version of you that ships release four is going to be glad you did. For artists who want to extend the system across a full catalog, build a brand asset library covers how to structure the Vault so the savings compound across 12 or more releases. If you want to see where character setup fits inside a full release timeline, the 21-day release week timeline shows exactly when to lock the character in the production sequence.
What kind of reference photos work best for an AI music video character
The reference photos you upload are the most important input to the character build. A strong set of references produces a reliable, consistent character across scenes. A weak set produces drift — a slightly different face from shot to shot that viewers notice even if they cannot name it.
What works:
Clean, well-lit headshots with the face centered and fully visible. No harsh shadows cutting across the face, no backlighting that silhouettes the head. Photos where the face, hair, and styling match what you want the AI to reproduce — the model carries the wardrobe and expression from your references into the generated character. Variety across the four slots: one close headshot, one mid-distance angle, and two profiles gives the model enough spatial information to reconstruct the face convincingly when the camera moves around the subject. Consistent styling across all references — if three photos show dark hair and one shows bleached hair, the model will average the two, which is usually worse than either.
What to avoid:
Group photos: the model does not know which person in the group is the character. Heavy filters: these distort face shape and color in ways that hurt likeness accuracy. Extreme angles (looking straight up, looking straight down): these give the model limited information about the actual face. Photos where the face is partially obscured by sunglasses, large hats, or masks in the headshot slot — all reduce the information available to anchor the likeness. Low resolution images: screenshots from social media or heavily compressed thumbnails often perform worse than the same photo at its original resolution.
The headshot is the single most important slot. If you only upload one reference, make it a clean, well-lit close-up with no obstructions, taken in conditions that match how you want the character to look on screen.
FAQ: Setting up an AI music video character in Echonos
What is Echonos Characters?
Echonos Characters is a feature inside Echonos that lets you save a named on-screen identity — a character — built from up to four reference photos. The character lives in your Vault and can be selected for any video generation. When you apply a character to a generation, the pipeline uses the reference photos to anchor the same face, body, and silhouette across every scene of the video and every future video on your account. It is the mechanism that keeps your catalog visually consistent without needing to re-brief the artist's appearance on every release.
Can I use the same AI character across multiple videos?
Yes. That is the entire point of the Characters feature. Once you save a character to your Vault, it is available as a selectable asset on every future Create flow. You pick the character before generating, and the pipeline applies the same reference identity across every scene. The character persists across separate projects, sessions, and releases — unlike reference image conditioning in general-purpose tools, which only works within a single conversation or project. Artists using Echonos regularly use the same character across an entire single cycle, EP, or album campaign without needing to rebuild the reference on each release.
How many personas can I save in my Vault?
There is no hard cap on the number of Characters you can save. Many solo artists keep a single persona that runs across an entire era, then create a second one when they pivot to a new release cycle. Managers and labels routinely run a separate persona per artist on the roster, which is the intended pattern for multi artist workflows. Old personas stay in the Vault even after you build a new one, so older releases keep working with the persona they were originally generated against.
Does creating or editing a persona cost credits?
No. Persona creation and edits do not consume credits. Credits are spent only on actual video generation, charged as flat fees per operation (a full Engine generation is 200 credits regardless of song length). That means you can build a persona, swap reference images, rename it, and rebuild it as often as you want before you ever run a generation. The credit decision happens at the generate step, not the persona step.
What kinds of reference images work best?
Strong personas come from references that are visually consistent with each other. The same person, the same general lighting, and the same wardrobe direction across all your reference slots produces a tighter lock than a mix of dressed up promo shots and casual phone selfies. The photos do not need to be professional, but they should agree with each other on the things you want the model to remember, especially face, silhouette, and style.
Can I use a non human character as a persona instead of myself?
Yes. The Characters builder accepts any visual identity, not just real people. Some artists save a styled illustrated character, a masked persona, or a recurring motif as the on screen identity for a project. The same locking logic applies: upload reference images of the character from the angles and contexts you want preserved, and the persona will reappear consistently across every video you generate.
Keep reading
Related Articles
Character Consistency in AI Music Videos: Why It Is the Foundation of Your Artist Brand in 2026
Consistent character AI keeps the same face and presence stable across every music video release. Here is what it means and how Echonos handles it with a saved Vault asset.
What Makes a Good Music Video: The Elements That Actually Hold Attention
What makes a good music video comes down to five repeatable elements: concept, consistency, pacing, a recognizable look, and how you generate it. Here's the breakdown.
Hip Hop Release Content: Lyric Videos, Character Shots, and Drop Centered Music Videos in 2026
How to make a rap music video and hip hop release content in 2026: character-forward visuals, lyric videos that travel, and drop-centered cuts.
Written by
Syed Ali
Founder & CEO
Former COO at Tabler App (1M+ users, $50K+ MRR, successful exit) and Data Science Consultant at Deloitte, Booz Allen Hamilton, and Accenture. Leads capital markets, investor relations, and corporate strategy at Echonos.