A Fiverr gig gallery is a stills row, not a freelancer talking head
The gig SKU is three to five labeled stills of the deliverable. Qwen holds the UI type. /for/freelancers is the person. A generated talking seller is a different ad and usually the wrong muted scan.
The gig SKU is three to five labeled stills of the deliverable. Qwen holds the UI type. /for/freelancers is the person. A generated talking seller is a different ad and usually the wrong muted scan.
A Fiverr buyer does not press play. They scan a row: package name, a crop of the file they will get, a label they can proof against the order page. That row is already stills. Motion is optional and usually wrong. A constructed seller explaining “why you should book me” is a personal-brand clip. It is not the gallery.
Three to five labeled stills of the deliverable
Marketplace galleries are judged as a grid. Buyers match type, not monologues. The frames that answer “is this the gig that delivers a logo pack / a captioned reel” are the stills: readable heading, true package names, a crop of the actual file. Inventing a prettier desk answers “does this look like a movie about freelancing.”
Quote the glyphs: the gig title, “Basic / Standard / Premium,” delivery days, a file-type line you will actually ship. Capture real screenshots when you have them. Generated type is for the graphic that is not in the file yet — a 9:16 crop the marketplace thumbnail will still read. Keep other people’s client faces out. A photoreal “happy founder reviewing my work” is a likeness you did not clear.
Three to five stills is the SKU. One talking clip is not a substitute for four missing frames. A mute scan will never hear the pitch. It will only see whether the type holds.
Qwen holds the UI type; Seedream holds the dense row
Lettering is a still job. Most photoreal models will give you a marketplace-shaped texture with word-shaped smears. That is not a gig title. Video models are a worse typesetter.
Qwen Image 3 Text to Image is the stills row when the letters have to read. Catalog facts: Qwen’s third-generation image model, strong prompt adherence, complex text rendering in Chinese and English, intelligent prompt rewriting on by default, 4 credits, 2K max, qualities 1K and 2K. No audio. No duration. Aspects include 1:1, 16:9, 9:16, and 21:9.
Write the exact strings. If the gig title or a package name must stay verbatim, say so in the prompt and inspect the still. Rewriting is a help for scene language, not a licence to paraphrase “Logo pack (3 concepts)” into “branding vibes.” Set 1K while the words are in draft. Set 2K when the gallery tile is the tile. There is no 4K chip on this row.
A full extras grid is denser than a three-word title. Seedream 5.0 Pro is ByteDance’s flagship text-to-image row: 7 credits, native text rendering in 14 languages, control over dense layouts and structured designs, qualities listed at 2K. Quote the schema. How many packages. Which extras. Proof at 100%.
Use Qwen when the hero is a heading plus one labeled crop. Use Seedream when the hero is the extras table. A real screenshot beats both. Text to image is the catalog door for that still.
Generate an image from text is the named still job: you describe the cover, the agent asks style, aspect, and count, then picks an image model — or uses the one you name. Priced per model, shown before you confirm. Name the row. “Qwen Image 3, gig title and these three package names, no people.” Stop. “Make me a freelancer talking about my gig” is the wrong object.
/for/freelancers is the person
Freelancers convert with process: a real work-in-progress screen recording, a client-call snippet with permission, a workspace tour. Three posts a week. 9:16. LinkedIn and Reels. The converting object is how you work. A Fiverr gallery that copies that playbook — talking creator, invented studio, a “hire me” generate — has used the wrong still. Give the process away on the /for page. Typeset the gallery here.
A screen recording of you designing is a LinkedIn asset. A labeled still of the deliverable is the gig. Mixing them is how a buyer lands on a talking face and never sees the package names. If a human must speak, that human is a real named seller on a different day, pointing at the approved stills.
A generated talking seller is the wrong muted scan
A cinematic “gig seller reacting to a five-star review” is talent you did not hire. The row already is the CTA. Prompted type in a video generate is a miss you cannot edit without regenerating the face. Marketplace buyers will not unmute to hear a pitch the gallery never showed.
Do not text-to-video a neon desk with a glowing “Order now” button. Do not lipsync a constructed twin holding “your” Fiverr handle. Do not replace four missing stills with one eight-second monologue. Change the package set with a new screenshot or a new Qwen still. Add a text overlay to a video burns one authored line if you later animate an approved still. Most listing weeks should stop at the stills.
Pause on the row. Three to five labeled frames, letters you can proof: this listing. A mouth on a stage: the freelancer /for.
FAQ
Can a talking video model typeset the gig title if I write it in the prompt?
Do not. Prompted type in a video generate is a miss you cannot edit, and the model will still invent a host around it. Generate the tiles as stills on Qwen Image 3 or Seedream 5.0 Pro, or keep screenshots of the real deliverable.
Qwen or Seedream for the gallery?
Qwen Image 3 is the lettering scout: 4 credits, 2K, English and Chinese type, prompt rewriting on by default. Seedream 5 Pro is the dense-layout row: 7 credits, 14-language type, structured designs. A heading plus one labeled crop is Qwen. A full extras table is Seedream. A real screenshot beats both.
Is a generated talking seller acceptable if we disclose they are AI?
Disclosure does not make them the gallery. The gig SKU is three to five labeled stills of the deliverable. A constructed host is a second product and a likeness you did not need. Keep the UI. Keep the labels. Leave the face off the generate.
Can I run the /for/freelancers playbook on the Fiverr listing?
Not as a substitute for the stills row. That page is process: a real WIP recording, a workspace tour, a personal-brand clip. This page is the muted marketplace scan. A talking seller is the freelancer audience. Qwen plus Seedream is the gig SKU.