Seedream 5.5
par ByteDance Seed
Skin, light and small type that hold up in a close-up.
Change one thing and the rest of the frame stays put.

Points clés
Natural skin and light
Pores, freckles and the way window light falls on a face, so close-ups of a singer read as photographed rather than rendered.
Edits that keep the frame
Ask for one change, like the light or a jacket, and the rest of the picture stays as it was. ByteDance says unedited regions hold at the pixel level for up to a hundred rounds.
Dense small text
Tracklists, credits and labels in small type. In our test a six-song tracklist came out spelled exactly as written.
Up to ten references
Character sheets and a location in one request, so the same person carries across scenes.
Transparent PNGs
It can draw on a transparent background, useful for cut-out props and titles laid over footage.
1K to 2K
Output at 1K, 1.5K or 2K in eight aspect ratios from 21:9 to 9:16. There is no 4K tier.
Caractéristiques techniques
Seedream 5.5
ByteDance Seed, released 9 November 2026
10 references
Public image URLs, up to 10 MB each
1K / 1.5K / 2K
No 4K tier
21:9 to 9:16
Eight aspect ratios
~25 s
A 2K frame with two references, measured
PNG or JPEG
Transparent backgrounds in PNG
0.4%
Severe text errors in ByteDance's own evaluation
100 rounds
Of editing with untouched regions kept, per ByteDance
Exemples de prompts
Golden-hour van
Four friends in a band squeezed into the front seats of a tour van at golden hour, laughing, sunlight flaring across their faces, dust on the dashboard
Modifier le promptWindow close-up
Close-up of a man with freckles and a short beard singing softly into a vintage ribbon microphone beside a rain-streaked window, soft overcast daylight on his skin
Modifier le promptChange only the light
Change only the light: it is now dusk, the window glows deep blue and a warm table lamp lights his face from the left. Keep everything else exactly the same
Modifier le promptTracklist
Top-down photo of the back of a vinyl sleeve on a wooden table, printed with a six-song tracklist in small serif type, a coffee cup and headphones beside it
Modifier le promptChapel rehearsal
A string quartet rehearsing in an empty whitewashed chapel, soft daylight from tall windows, wooden chairs, sheet music on stands
Modifier le promptSuburban dusk
Two teenagers riding bikes down a suburban street at dusk, sprinklers on the lawns, streetlights just coming on, slight motion blur
Modifier le promptCas d’usage

Performance close-ups
A singer under one spotlight, close enough to see the skin, for the shots where the face carries the song.

Band and tour moments
Warm, candid frames of the band between the songs, the kind a documentary-style video cuts to.

Record sleeves and credits
Album backs, tracklists and title cards where every small word has to be right.

Quiet, wide rooms
Calm interiors in soft daylight for slower songs and string arrangements.
Présentation
Seedream 5.5 is ByteDance Seed's image model, released on 9 November 2026. It draws from a prompt or edits from up to ten reference images, at 1K, 1.5K or 2K. ByteDance pitches it on three things: skin and light that look photographed, edits that leave the rest of the frame untouched, and small, dense text that stays legible.
Edits that leave the frame alone
Most image models redraw the whole picture when you ask for one change. Seedream 5.5 is built to change only what you name. The two window close-ups on this page are one edit apart: the second asked only for dusk light and a warm lamp, and the face, the microphone and the framing stayed where they were.
For a music video that means a keyframe can be corrected, such as the time of day, a jacket or a prop, without losing the parts that were already right.
Small text
ByteDance reports severe text corruption in 0.4% of its evaluation images and minor errors in 1.7%. In our test the full six-song tracklist on the record sleeve came out spelled exactly as written. Dimly still keeps keyframes with lyrics in the frame on GPT Image 2.6, which is what its lettering checks are tuned for.
Seedream 5.5 and Dimly
Every Dimly music video starts as images: character sheets, look previews and one keyframe per shot, which a video model then animates. You don't pick the image model in the studio. Dimly's director draws with GPT Image 2.6 by default and can be switched to Seedream 5.5. It costs about the same per keyframe as GPT Image 2.6, because each reference image adds to the price.
Tarification simple
Préparez votre vidéo gratuitement, puis choisissez le plan adapté à votre calendrier de sortie.
Starter
Un mois régulier de Chorus Cuts, ou une chanson plus courte.
- 600 crédits, valables 30 jours
- ≈ 8 Chorus Cuts (15 s)
- ou une chanson allant jusqu’à 2:20
- Tous les modèles et toutes les résolutions · 3 rendus à la fois
Pro
Une chanson complète par mois avec de la marge pour refaire des prises, et plus de crédits quand une sortie en a besoin.
Crédits par mois
- 1 000 crédits, valables 30 jours
- ≈ 14 Chorus Cuts (15 s)
- ou 1 chanson complète (3:45)
- Tous les modèles et toutes les résolutions · 4 rendus à la fois
Creator
Plusieurs chansons complètes par mois, jusqu’en 4K.
- 5 000 crédits, valables 30 jours
- ≈ 72 Chorus Cuts (15 s)
- ou ≈ 5 chansons complètes (3:45)
- Tous les modèles et toutes les résolutions · 8 rendus à la fois
Chanson complète, Chorus Cut, vidéo lyrics ou visualiseur · Exports sans filigrane sur tous les forfaits · Les droits d’export dépendent de vos autorisations sur la source et des conditions des fournisseurs
Packs de crédits
Payez une fois, sans abonnement. Pensé pour un effet vidéo IA ou une chanson ; si vous comptez en créer plusieurs, un plan ci-dessus est plus avantageux.
À l’unité
Une vidéo d’effet IA, comme Hotel Lobby ou AI Rap Duo.
- 250 crédits, valables 1 an
- 1 vidéo d’effet IA (Hotel Lobby, AI Rap Duo)
- ou ≈ 3 Chorus Cuts (15 s)
- Les exports portent le filigrane Dimly
- Sans abonnement, jamais renouvelé
Chanson
Un clip complet (une chanson jusqu’à 3:45), ou quatre effets IA.
- 1 000 crédits, valables 1 an
- 4 vidéos d’effet IA (Hotel Lobby, AI Rap Duo)
- ou 1 chanson complète (3:45)
- Exports sans filigrane
- Sans abonnement, jamais renouvelé
Studio
Une chanson complète en 2K avec de la marge pour refaire des prises, ou deux chansons.
- 2 000 crédits, valables 1 an
- 8 vidéos d’effet IA (Hotel Lobby, AI Rap Duo)
- ou ≈ 2 chansons complètes (3:45)
- Exports sans filigrane
- Sans abonnement, jamais renouvelé
Les estimations utilisent MiniMax H3 en 768P avec des plans de 5 secondes (un Chorus Cut de 15 secondes compte pour trois), une keyframe par plan, deux planches de référence et aucune reprise ; une chanson complète dure 3:45, la durée typique des chansons qu’on nous apporte. Les résolutions supérieures et les autres modèles coûtent plus cher par vidéo. Vous voyez le devis exact avant tout rendu, et les rendus échoués sont remboursés.
FAQ
ByteDance Seed's image generation and editing model, released on 9 November 2026. It takes a prompt and up to ten reference images and draws one image at 1K, 1.5K or 2K.
Not as a choice in the studio: you don't pick the image model. Dimly's director draws character sheets, look previews and keyframes with GPT Image 2.6 by default and can be set to use Seedream 5.5 instead. Keyframes with lyrics in the frame stay on GPT Image either way.
No. It outputs 1K, 1.5K or 2K; at 9:16, 2K came out at 1584 × 2816 in our test.
One to ten, passed as public image URLs of up to 10 MB each, in JPEG, PNG or WebP.
Yes. Ask for a transparent background with PNG output and it draws the subject with no background.
Dimly generated them with Seedream 5.5 at 2K on 15 November 2026, through the same Kie route the studio uses, and downscaled them for the web. The dusk window close-up is an edit of the daylight one, and the club stage shot uses the same portrait as its reference.
Autres modèles
Voir tous les modèlesMiniMax H3
MiniMax
Reference-driven video with native stereo sound. Your track drives the picture, 768P up to 4K.
Seedance 2.5
ByteDance
First-and-last-frame moves up to thirty seconds, audio-driven at 480p/720p — your song muxed back intact.
Kling 4.0
Kuaishou
Kuaishou's long-take video model: up to thirty seconds in one pass, ten keyframes and Omni Reference with up to fifteen inputs.
MiniMax H3 Image
MiniMax
MiniMax's open-weights image model: up to sixteen reference images, native 4K and Chinese or English lettering, in about twenty seconds.




