01
Ad hooks you can test in batches
Write six versions of the same opening beat, hold the product and brand colours fixed, and vary only the action or the camera move. The cheaper tiers make that volume affordable before you commit to a hero cut.
Text to Video, Free:1080P Seedance, No Watermark
Write the scene and Seedance shoots it. Subject, action, camera move, and light all follow the words you type, with no queue and no watermark on what you download. New accounts start with free credits.
Real output
One prompt, one finished shot
Nothing was attached to this generation: no reference image, no editing pass. The prompt below is the entire input, and the clip beside it is what came back.
Prompt
A macro FPV chase shot follows a striped bee skimming inches above a busy highway at sunset, weaving between speeding cars, tires, lane markers, dust, and road debris. The low camera races across the asphalt with intense motion blur and warm golden-hour light. The bee narrowly escapes traffic as a huge explosion erupts behind it, flames and smoke swallowing the road, cinematic high-speed action.
The full prompt, exactly as it was sent.
Result
Seedance output, downloaded without a watermark.
Use cases
What people write from scratch
Text to video is the fastest way to test an idea that does not exist as footage yet. These are the jobs it takes on most often here.
02
Storyboards and previz
Turn a written shot list into moving frames, so a director, a client, or a teammate reacts to pacing and blocking instead of a static board.
03
Social clips at platform ratios
Choose the aspect ratio first and write for it. Vertical hooks for short-form feeds, wide frames for pre-roll, and no reframing pass afterwards.
04
B-roll you cannot shoot
Establishing shots, weather, crowds, macro detail, aerials: the footage that would otherwise need a location, a permit, or a drone.
How it works
How text to video works here
There is no mode to pick. The composer reads your input and routes itself, so attaching nothing keeps the job on text to video.
01
Describe the shot, not the topic
Name the subject, what it does, where the camera sits, and how it moves. Lens and lighting words such as golden hour, macro, or handheld land far harder than adjectives such as beautiful.
02
Choose a model and a length
Seedance 2.5 runs up to 30 seconds. Seedance 2.0, Fast, and Mini trade ceiling for cost, and Veo 3.1, Sora 2, Kling 3.0, and MiniMax H3 sit in the same picker.
03
Generate, review, keep the take
Finished clips land in your asset library at full quality with no watermark, ready to download or to attach back in as a reference.
Jobs people arrive with
Four text-to-video jobs, answered directly
These are the questions people bring to an empty prompt box, worded the way they ask them. Each answer says whether it works here, on which model and with what limit, then gives the steps and a prompt you can paste as written.
01
Make an AI video with sound and spoken dialogue from text
Yes. Seedance 2.5, Seedance 2.0, 2.0 Fast, 2.0 Mini and 1.5 Pro, Kling 3.0, Kling O3 and Wan 3.0 all have an audio switch here, and with it on the clip comes back with ambience, effects and voices generated together with the picture. Spoken lines work when the words are in quotes and the speaker is named. On the Seedance models the switch does not change the credit cost, and Seedance 2.0 Mini, the model that runs on free credits, has it too. Veo 3.1, Sora 2 and MiniMax H3 show no audio option in this workspace.
How to do it
- Pick a model with the audio switch and leave it on.
- Describe the picture first, then the sound: the room tone, one or two effects, any music.
- Put each spoken line in quotes and name who says it; one or two short lines fit a clip under ten seconds.
- Generate, and switch audio off on a rerun if you want a silent clip for your own mix.
Prompt
Night market in the rain, close-up on a street cook flipping skewers over a charcoal grill, sparks lifting into the steam. Handheld, shallow focus. She glances up at the camera and says, "Two more minutes, they are almost ready." Rain drumming on the canvas awning, fat sizzling on the coals, chatter behind her.
Settings
- Model: Seedance 2.0 Mini on free credits · Seedance 2.5 for longer takes and 1080p
- Duration: 8 seconds · Aspect ratio: 16:9
- Audio: on; on the Seedance models it costs the same as off
Prompt recipe: no clip attached, paste it and run it.
02
Generate a 30-second video in one take, with several shots
Seedance 2.5 renders 4 to 30 seconds in a single run, and so does Wan 3.0; most other models here stop between 8 and 15 seconds. For a multi-shot take, give each shot its own time range in the prompt, such as 0–10 s, 10–20 s and 20–30 s, and describe the camera inside each range. Thirty seconds is the length people choose most often for Seedance 2.5 text to video on this site, and it needs a paid plan or a credit pack; the cost table below prices a 15-second 1080p run on every plan.
How to do it
- Choose Seedance 2.5 and set the duration to 30 seconds.
- Split the prompt into time ranges, one shot per range, and describe the subject and the light the same way in each.
- Say whether shots cut or flow into one continuous move, so the model does not have to guess.
- If one shot misbehaves, shorten or simplify that range instead of rewriting the whole prompt.
Prompt
Launch film for a handmade ceramic mug, three shots. 0–10 s: macro of wet clay spinning on a potter’s wheel, hands shaping the rim, soft window light. 10–20 s: cut to the glazed mug lifted out of a kiln, orange glow, slow push-in. 20–30 s: cut to a kitchen table in the morning, coffee poured into the mug, steam rising, the camera settles on it. Warm film look, the same mug colour in every shot.
Settings
- Model: Seedance 2.5, paid plan or credit pack required
- Duration: 30 seconds · Aspect ratio: 16:9
- Resolution: 720p on Lite, 1080p on Pro, Premium and credit packs · Audio: on
Prompt recipe: no clip attached, paste it and run it.
03
Make a vertical 9:16 video for TikTok, Reels or Shorts from a prompt
Yes, on every model in the table above: set the aspect ratio to 9:16 before you generate and write the shot for a tall frame, with the subject in the middle and the action moving toward or away from the camera rather than across it. Seedance 2.0 Mini does it on free credits at up to 720p, which is enough to test a hook; Seedance 2.5 and 2.0 reach 1080p on Pro and Premium. The clip is generated vertical, so there is no crop step and nothing is cut off the sides.
How to do it
- Set 9:16 first, then write the prompt for that frame.
- Keep the subject and the important action in the middle of the frame, clear of where captions and buttons sit.
- Put the hook in the first two seconds: one clear action and one camera move.
- Generate three or four variants that change only the opening action, and keep the one that reads fastest.
Prompt
Vertical 9:16. A skateboarder rolls straight toward a low camera on an empty rooftop at sunset, kicks the board up into her hand right in front of the lens and grins. Low angle, wide lens, golden backlight, city skyline behind her, no text on screen.
Settings
- Model: Seedance 2.0 Mini on free credits (up to 720p) · Seedance 2.5 or 2.0 for 1080p
- Duration: 5 seconds · Aspect ratio: 9:16
- Audio: on when the hook depends on sound, off for a silent loop
Prompt recipe: no clip attached, paste it and run it.
04
Keep the same character across several AI videos
Words alone drift: the same description gives a slightly different face on every run. The dependable fix is reference images. Attach stills of the character and, on the Seedance models, the prompt stays the scene while the pictures hold the identity: up to 30 images on Seedance 2.5 and nine on Seedance 2.0, Fast and Mini. Name each image in the prompt, such as “the woman in image 1”, and reuse the same set for every clip in the series. Human faces are supported as references: of 6,214 image jobs run here between September 15 and October 14, 2026, only 19 were refused for showing a real person.
How to do it
- Make or choose two or three clean stills of the character: front, three-quarter and full body, on a plain background.
- Attach them to the composer; on the Seedance models they become a reference group rather than an opening frame.
- Refer to them by number in the prompt and describe only the action, the place and the camera.
- Keep the stills and the wording of the character identical across the series; change only the scene.
Prompt
The woman in image 1, wearing the yellow raincoat from image 2, runs up a flight of stone steps in a narrow old-town street at dusk, looks back over her shoulder and laughs. Tracking shot from behind, then she turns to the camera. Wet cobbles, warm street lamps.
Settings
- Model: Seedance 2.5 (up to 30 reference images) or Seedance 2.0 (up to 9)
- Inputs: the prompt plus two or three reference stills of the character
- Duration: 8 seconds · Resolution: 720p on Lite, 1080p on Pro and Premium
Prompt recipe: no clip attached, paste it and run it.
What you can generate
What text to video can generate here
Every row is read from the model registry the composer enforces, not from a brochure. The resolution ceiling, the duration range, the audio toggle and the reference limits are the same values the workspace applies when you press generate, and the table is rebuilt whenever the registry changes.
| Model | Max resolution | Clip length | Audio | Inputs | On free credits | Starting cost |
|---|---|---|---|---|---|---|
| Seedance 2.5 | 1080p | 4–30s | On/off toggle | Prompt only · Up to 30 reference images | Paid plan or pack | From 12 credits (4s · 480p) |
| Seedance 2.0 | 1080p | 4–15s | On/off toggle | Prompt only · Up to 9 reference images | Paid plan or pack | From 9 credits (4s · 480p) |
| Seedance 2.0 Fast | 1080p | 4–15s | On/off toggle | Prompt only · Up to 9 reference images | Paid plan or pack | From 7 credits (4s · 480p) |
| Seedance 2.0 Mini | 720p | 4–15s | On/off toggle | Prompt only · Up to 9 reference images | Yes | From 5 credits (4s · 480p) |
| Seedance 1.5 Pro | 1080p | 4–12s | On/off toggle | Prompt only | Yes | From 2 credits (4s · 480p) |
| SeeDance 1 Pro Fast | 1080p | 2–12s | No audio option | Prompt only | Yes | From 1 credit (2s · 480p) |
| Kling O3 | 1080p | 4–15s | On/off toggle | Prompt only | Paid plan or pack | From 10 credits (4s · 720p) |
| Kling 3.0 | 1080p | 4–15s | On/off toggle | Prompt only | Paid plan or pack | From 10 credits (4s · 720p) |
| Veo 3.1 Fast | 4K | 4–8s | No audio option | Prompt only | Paid plan or pack | From 12 credits (4s · 720p) |
| Veo 3.1 Pro | 1080p | 4–8s | No audio option | Prompt only | Paid plan or pack | From 32 credits (4s · 720p) |
| Sora 2 | Model default | 4–12s | No audio option | Prompt only | Paid plan or pack | From 8 credits (4s) |
| Sora 2 Pro | 1080p | 4–12s | No audio option | Prompt only | Paid plan or pack | From 24 credits (4s · 720p) |
| MiniMax H3 | 2K | 5–15s | No audio option | Prompt only · Up to 9 reference images | Paid plan or pack | From 23 credits (5s · 2K) |
| Wan 3.0 | 1080p | 2–30s | On/off toggle | Prompt only · Up to 10 reference images | Paid plan or pack | From 2 credits (2s · 480p) |
- Free credits cover the three entry models: Seedance 2.0 Mini, Seedance 1.5 Pro and SeeDance 1 Pro Fast. Any paid plan unlocks every row in the table, and one-time credit packs unlock models by tier, so a Seedance 2.5, Kling 3.0 or Veo 3.1 job needs a plan or a pack before it will run.
- Clip length is per generation. Seedance 2.5 and Wan 3.0 reach 30 seconds in one run; most other models stop between 8 and 15 seconds, and a longer sequence is built by extending or by cutting several takes together.
- Audio means a soundtrack generated together with the picture. Models with an on/off toggle can return either a scored clip or silent video; the others expose no audio option in this workspace.
- Starting cost is the catalogue rate for the shortest, lowest-resolution clip the model offers, the same figure the comparison table on the pricing page quotes. Paid plans charge below catalogue, longer or sharper clips cost more, and the composer prices the exact job before you commit.
Resolution cap by plan
The plan sets the ceiling and the model sets what is possible under it. A 4K-capable model still renders at 720p on free credits, and Seedance tops out at 1080p on every plan.
| Plan | Seedance models | Other models |
|---|---|---|
| Free credits | 720p | 720p |
| Lite | 720p | 1080p |
| Pro | 1080p | 2K |
| Premium | 1080p | 4K |
Last updated
Cost in dollars
What a text-to-video clip costs in dollars
Credits are what the composer charges; this table turns four common jobs into dollars at each plan’s price per credit. The credits come from the same calculation the generator bills with, plan discounts included, and the prices from the pricing page, so the two move together.
| Plan and price | Seedance 2.0 Mini5 s · 720p | Seedance 2.010 s · 720p | Seedance 2.510 s · 720p | Seedance 2.515 s · 1080p |
|---|---|---|---|---|
| Lite, monthly$14.90/mo · 100 credits/mo | $1.7912 credits | $5.3636 credits | $8.2055 credits | Not on this plan |
| Lite, billed yearly$13/mo · 100 credits/mo | $1.5612 credits | $4.6836 credits | $7.1555 credits | Not on this plan |
| Pro, monthly$29/mo · 200 credits/mo | $1.319 credits | $4.0628 credits | $6.2443 credits | $23.06159 credits |
| Pro, billed yearly$14.50/mo · 200 credits/mo | $0.659 credits | $2.0328 credits | $3.1243 credits | $11.53159 credits |
| Premium, monthly$149/mo · 1,300 credits/mo | $1.039 credits | $3.2128 credits | $4.9343 credits | $17.88156 credits |
| Premium, billed yearly$99/mo · 1,300 credits/mo | $0.699 credits | $2.1328 credits | $3.2743 credits | $11.88156 credits |
| Starter credit pack$29.90 one-time · 200 credits | $1.9413 credits | $5.9840 credits | $9.1261 credits | $33.94227 credits |
- Each dollar figure is the plan’s price per credit times the credits that job costs on that plan, for someone subscribing today. The credits a job costs differ by plan, which is why every row is computed on its own.
- Sign-up credits are not in the table: they cover a first finished clip on the entry models, Seedance 2.0 Mini, 1.5 Pro and SeeDance 1 Pro Fast, at up to 720p, with no card on file.
- Lite caps Seedance at 720p, so the 1080p column is empty on Lite. Turning audio on does not change the Seedance price, and a failed or refused generation is not charged.
- The composer shows the exact credit cost of the job you have set up before you press generate; if it ever disagrees with this table, the composer is right.
Measured here
How long text to video takes here, measured
Timed on this site from the moment a job is submitted to the moment the finished clip is ready, across more than 4,700 text-to-video jobs that finished between September 15 and October 14, 2026. These are real waits, not progress-bar estimates, and every row has at least 30 finished jobs behind it.
| Model | Setting picked most often | Median wait | 3 in 4 done within |
|---|---|---|---|
| Seedance 2.0 Mini | 4 s · 480p | 2 min 22 s | 2 min 52 s |
| Seedance 2.5 | 30 s · 720p | 4 min 42 s | 6 min 30 s |
| Seedance 1.5 Pro | 5 s · 480p | 1 min 20 s | 1 min 44 s |
| SeeDance 1 Pro Fast | 5 s · 480p | 51 s | 1 min 6 s |
| Kling 3.0 | 5 s · 720p | 2 min 36 s | 3 min 10 s |
| Seedance 2.0 | 10 s · 720p | 6 min 53 s | 9 min 52 s |
- The wait covers everything between pressing generate and the clip appearing in your library: the queue at the model provider, rendering and delivery. Longer clips and 1080p take noticeably longer than the setting shown in each row.
- Jobs started in unlimited mode, which can wait in a queue by design, and draft previews are left out, so the figures describe an ordinary credit-billed run.
- Free credits and Lite run one job at a time, Pro runs four in parallel and Premium eight, so on a paid plan several of these waits overlap.
Worked examples
Three text-to-video jobs, end to end
The first example is the real clip from the demo above with its unedited prompt. The other two are prompt recipes written against the same models and limits, so you can paste them in and compare your result with the brief.
01
Macro FPV bee chase
Before
Nothing but the sentence below. No reference image, no source clip, no style preset: the composer read an empty attachment tray and routed the job to text to video.
After
A low, fast tracking shot over sunset asphalt with heavy motion blur, cars and tyres sweeping past the bee, and an explosion filling the frame behind it, all staged from the words alone. This is the clip that plays in the demo section above.
Prompt
A macro FPV chase shot follows a striped bee skimming inches above a busy highway at sunset, weaving between speeding cars, tires, lane markers, dust, and road debris. The low camera races across the asphalt with intense motion blur and warm golden-hour light. The bee narrowly escapes traffic as a huge explosion erupts behind it, flames and smoke swallowing the road, cinematic high-speed action.
Settings
- Model: Seedance 2.0
- Inputs: prompt only, nothing attached
- Entry: text to video, inferred from the empty tray
02
Vertical product hook on free credits
Before
A brief for a 9:16 social hook with no footage to start from: a matte black insulated bottle on wet slate, and the wish to test the opening beat before anyone books a studio.
After
A six-second vertical clip: the bottle standing in shallow water as ripples roll toward camera, a slow low-angle push-in, condensation catching a hard side light, no text and no logo. Change one clause per take to compare openings.
Prompt
A matte black insulated bottle stands on wet dark slate, shallow water rippling toward camera. Slow push-in from a low angle, hard side light raking across condensation, deep shadows, vertical 9:16, no text, no logo.
Settings
- Model: Seedance 2.0 Mini, available on free credits
- Duration: 6 seconds · Aspect ratio: 9:16
- Resolution: 720p, the Mini ceiling and the free-credit cap
- Audio: off
Prompt recipe: no clip attached, paste it and run it.
03
Thirty-second dawn sequence on Seedance 2.5
Before
A shot list for a short-film opening that would otherwise need a drone, a coastal location and a dawn call time: a fishing village waking up, told in three connected camera moves.
After
One continuous 30-second clip: a wide aerial drift over the harbour at first light, a descent into a slow dolly along the quay as shutters open, and a final tilt from wet cobbles to the church tower, with generated ambient sound under it.
Prompt
A coastal fishing village at dawn, three connected camera moves. 0–10s: slow aerial drift over a small harbour, boats rocking, mist lifting off the water, pale gold light. 10–20s: descend to a low dolly along the stone quay as wooden shutters open and a baker carries trays past camera. 20–30s: tilt up from wet cobblestones to the church tower as the first bell rings, gulls crossing the frame. Anamorphic look, soft haze, ambient harbour sound.
Settings
- Model: Seedance 2.5, paid plan or credit pack required
- Duration: 30 seconds · Aspect ratio: 21:9
- Resolution: 1080p on Pro or Premium, 720p on Lite
- Audio: on
Prompt recipe: no clip attached, paste it and run it.
Limits
Where text to video falls short, and why a job gets refused
Knowing the failure modes saves credits and retries. The refusal figures are counted from the jobs run on this site between September 15 and October 14, 2026.
01
Brands, logos and famous characters get refused
Most failed jobs here are not technical errors: more than eight in ten were stopped by the model’s content check, and the reason it names most often is copyrighted or trademarked material, such as a brand name, a logo or a well-known film, game or cartoon character. Describe the look instead of naming the brand or the character. A refused job is not charged.
02
Public figures by name
Faces are fine, and reference photos of people are accepted: refusals for showing a real person were about three in a thousand image jobs here. Naming a celebrity or another public figure in the prompt is a different matter and can be refused, so describe the look you want instead of the person.
03
Text inside the picture
Signs, labels, captions and logos drawn by the model often come out misspelled or warped, especially when they are small or moving. Keep text out of the prompt and add titles in your editor afterwards.
04
Fast hands and crowded action
Fingers, quick hand movements, tangled limbs and dense crowds are where artefacts show up first. One subject with one clear action, and a camera that does not whip around, gives the cleanest result.
05
Long takes drift
The longer the clip, the more a face, an outfit or a prop can change between the first and the last second. Split a long idea into timed shots on Seedance 2.5, or attach reference images so the identity is held by pictures rather than adjectives.
06
Timing is a strong hint, not a frame-accurate cue
Time ranges and shot order in a prompt are followed closely but not to the frame. When a beat has to land on an exact moment, such as a music hit, generate shorter clips and cut them together in an editor.
Why Seedance refuses prompts, counted on 11,168 jobs, with rewrites
Questions
Text to video questions
- Is text to video really free, and do I need a credit card?
- Yes to the first, no to the second. New accounts receive free credits with no card on file, and the grant is sized for a first finished clip: it allows two generation attempts and one completed video before a plan or a one-time credit pack is needed. On free credits the picker offers the entry models, Seedance 2.0 Mini, Seedance 1.5 Pro and SeeDance 1 Pro Fast, at up to 720p. A card only comes into play if you later choose a plan or a pack, and subscriptions can be cancelled at any time.
- Will there be a watermark on the free video?
- No. Nothing rendered here carries a watermark, on free credits or on a paid plan, so a clip can go straight into an edit. The file you download is the file the model returned, there is no overlay in the player, and there is no separate paid step to remove anything afterwards.
- Which models can I use for text to video?
- Every model in the table above accepts a prompt with nothing attached: Seedance 2.5, Seedance 2.0, 2.0 Fast, 2.0 Mini, 1.5 Pro and SeeDance 1 Pro Fast, plus Kling 3.0 and Kling O3, Veo 3.1 Fast and Pro, Sora 2 and Sora 2 Pro, MiniMax H3 and Wan 3.0. Kling 3.0 Motion Control and Kling 2.6 Motion Control are the two video models left out, because they need a still and a reference clip to work from. Free credits cover the three entry models; any paid plan opens the rest, and credit packs open them by tier.
- How long and how sharp can a clip be?
- Seedance 2.5 generates 4 to 30 seconds per run, Seedance 2.0, Fast and Mini go to 15 seconds, Wan 3.0 also reaches 30, and Veo 3.1 stops at 8. Resolution has two ceilings: the model’s own and the plan’s. Seedance tops out at 1080p on every tier, and the plan cap is 720p on free credits and Lite, 1080p on Pro and Premium. Other models render at up to 1080p on Lite, 2K on Pro and 4K on Premium where the model allows it; MiniMax H3 only outputs 2K, and Veo 3.1 Fast is the one model that offers 4K.
- How many credits does one clip cost?
- The starting-cost column shows each model’s catalogue rate at its smallest spec, before plan discounts, which is the same figure the comparison table on the pricing page quotes. Longer clips and higher resolutions raise the number, and every paid plan charges fewer credits per generation than the catalogue rate. The composer shows the exact cost of the job you have set up before you press generate, and a failed generation is not charged. For dollars, the cost table above converts four common jobs at each plan’s price per credit.
- Is seeddance.io the official ByteDance site for Seedance?
- No. SeedVideo AI at seeddance.io is an independent platform, not a ByteDance property. It runs the official Seedance model API alongside Kling, Veo, Sora, MiniMax and Wan, so the Seedance output you get here is produced by the model ByteDance publishes, but the workspace, the credits and the plans are ours. Model names belong to their respective owners, and nothing on this site is endorsed by them.
- Can I use the videos commercially?
- Commercial use is a plan feature. It is included on Pro, Premium and the annual Lite plan, and not on free credits or the monthly Lite plan. The output file is the same either way; the licence is what differs. You also remain responsible for what you write and attach, so keep brand names, real people and other people’s work out of prompts unless you hold the rights.
- How long does a generation take?
- A few minutes. Measured on this site over the 30 days to October 14, 2026, the median text-to-video job finished in 51 s on SeeDance 1 Pro Fast, 1 min 20 s on Seedance 1.5 Pro, 2 min 22 s on Seedance 2.0 Mini, 2 min 36 s on Kling 3.0, 4 min 42 s on Seedance 2.5 and 6 min 53 s on Seedance 2.0, each at the setting people pick most often; the measured table above has the full figures. Longer clips, 1080p and busy periods stretch them, and the workspace shows an “about N min left” estimate while it works. Free credits and Lite run one generation at a time; Pro runs four in parallel and Premium eight.
- Why was my text-to-video generation refused?
- Almost always because of the model’s content check rather than a fault: more than eight in ten failed jobs on this site between September 15 and October 14, 2026 were refusals. The most common triggers are brand names, logos and well-known characters, and wording the check reads as violent, sexual or otherwise sensitive. Describe the look instead of naming the brand or the character and run it again. A refused job is not charged.
- How do I keep a character or product consistent across clips?
- Write the fixed details the same way every time: name the subject, the wardrobe or packaging, the colour palette and the lens, and change only the action or the camera move between prompts. For tighter control, switch entry. Attach a still and the job becomes image to video, or attach a reference group on the Seedance models, up to 30 images on Seedance 2.5 and nine on Seedance 2.0, and the model holds identity from pictures rather than from adjectives. Finished clips can be attached back in as references for the next pass.
- Text to video or image to video: which should I use?
- Start from text when the shot does not exist yet and you want to explore: six versions of a hook, a previz pass, B-roll you cannot film. Start from an image when the look is already decided, because a still pins identity, palette and composition far more tightly than words can. A common workflow is to make the first frame as a still with an image model, then animate it, so the writing happens once and the motion is directed separately.
Choosing an entry
Text, image or video: which entry fits
The three scene pages share one composer. What you attach decides the entry, and the model you pick decides the ceiling.
Text is the right entry when nothing exists yet and the fastest way forward is to describe it. When you already own the frame, a packshot, a poster, a portrait, switch to image to video, because a still pins identity, palette and composition more tightly than any adjective. When the footage exists but one thing in it has to change, use video to video, where the instruction describes the difference rather than the scene.
The model choice is separate from the entry. Seedance 2.5 is the pick for a single 30-second take, multi-shot structure and reference-heavy jobs; Seedance 2.0 Mini is the pick for cheap volume on free credits. Every model in the table accepts a plain prompt, so the same brief can be run through Kling 3.0, Veo 3.1 and Sora 2 without rewriting it, and the results compared side by side in the asset library.
Keep going
Where to go next
Text is one of three ways in. Pick the entry that matches what you already have.
Free credits
Write it, then watch it
Open the generator, paste a scene, and keep the first take you like. No queue, no watermark, free credits on a new account.
Start from text