Choose quality, models and presets

Set Draft, Balanced or High quality, Step by step, presets and per-kind models, and use a planning budget and cost check before you send.

Updated

Set quality and models for your next message

  1. In New chat or an open conversation, select the quality button beside Send. It shows Balanced unless you or your defaults changed it.
  2. The Quality & cost menu opens. Choose Draft, Balanced or High quality.
  3. Leave Step by step off unless you have read Step by step.
  4. In a new chat, choose the Agent if more than one is offered. See The agent.
  5. To start from a saved set of models, open Saved preset and pick one.
  6. To pick a model for one kind of media, select Show model choices, then the row for Images, Video, Voice or Music & sound.
  7. Close the menu and send your message.

Your choices are saved with the draft. The menu says Saved with this draft. when they are stored. They apply to your next message: the menu reads Applies to your next message., or Model choices apply to your next run. while the agent is working.

Check that it worked

  • The quality button shows the level you chose, followed by · Step by step when that switch is on.
  • Any model you picked that differs from the default preset's, and an agent other than the default, appears as a chip above your brief. Select the chip to reopen the menu, or its close button to go back to the default for that kind.
  • While a reply is running, those chips are labeled Next run, because the running reply keeps the choices it started with.

Quality levels

LevelWhat the menu saysWhat changes
DraftFast exploration. Usually costs less.The agent is asked for fewer takes and lighter checks. Images left on Auto start on Gemini 3.1 Flash Image.
BalancedQuality and cost in balance.The default. Nothing is added to your message.
High qualityMore refinement. Usually costs more.The agent is asked for more takes, closer review and fixes.

Quality is direction for the agent, not a fixed setting. It doesn't set a price or guarantee a number of takes. To put a ceiling on spending, use the planning budget.

Step by step

Step by step is a switch in the same menu. The menu describes it as: "On: Wavemaker proposes each step and waits for your approval. Off: it carries the work through on its own."

  • On. Before each consequential step, such as a generation or an edit, the agent asks for approval and waits. Reading, searching and checking a price run without asking. While it waits, the conversation shows Needs you and "This request needs attention. Your work is saved."
  • Off. The agent carries the work through on its own, within your balance and any limit you set.

A step that isn't approved within 15 minutes counts as a decline: the agent doesn't run it and asks you what to do instead. Like the other choices in this menu, the switch applies to your next message. When you're signed in to a shared workspace account, the workspace's General settings page has Review a plan before generation, which starts new chats with Step by step on.

When the agent asks you a question as a choice card, in either mode, choose an option or reply in your own words.

Presets

A preset is a saved set of models, one per kind of media, plus a quality level. Saved preset in the quality menu lists two groups:

  • Standard presets: sets Wavemaker maintains. The recommended one is marked · Recommended.
  • Your presets: sets you saved.

Choosing a preset sets the quality and the models in one step. If you already picked a voice or music model yourself, choosing a standard preset keeps that pick. When your choices don't match any preset, the list shows Custom choices.

Standard presets

The standard presets offered today:

PresetQualityModelsMenu note
Standard · H3 + Gemini ProBalancedSelf-hosted H3 video, Gemini 3 Pro Image stills, ElevenLabs speechBalanced spend · self-hosted video. The default for new conversations.
Draft · H3 + Gemini FlashDraftSelf-hosted H3 video, Gemini 3.1 Flash Image stills, ElevenLabs speechLower-cost exploration
Gemini Omni + Gemini ProHigh qualityGemini Omni 1.1 Flash video, Gemini 3 Pro Image stills, ElevenLabs speechPartner video · higher spend
Kling 3.0 + Gemini ProHigh qualityKling 3.0 video, Gemini 3 Pro Image stills, ElevenLabs speechPartner video · higher spend
Permissive · self-hostedBalancedSelf-hosted Krea 2 stills and H3 video, for adult (18+) workSelf-hosted runtime

Wavemaker can update the standard presets. The quality menu always shows the current list, and the line under Quality & cost describes the preset your choices match.

Save, delete or reset a preset

  1. Set the quality and models you want.
  2. Select Save preset, type a name under Preset name (up to 60 characters), and select Save.
  3. To delete one of your presets, choose it under Saved preset and select the trash button, Delete preset.

Saving with a name you already use updates that preset. You can save up to 30 presets. Reset defaults returns the menu to Balanced with every model on its default and clears your default preset.

Your default preset and quality

Open Settings › General to choose what new conversations start from.

  • Default preset: pick a standard preset, one of yours, or No preset · use creative quality, then select Save default preset. The page confirms Default saved for new conversations.
  • Create or edit your presets: start from an existing preset under Start from preset, set Your preset name, Preset quality and a model for each kind, then select Save my preset. Delete my preset removes the one you started from.
  • Creative defaults: Default creative quality applies when no default preset is selected. Select Save creative quality.

Drafts you already started, and work already queued or running, keep their choices.

Choose a model for one kind of media

Select Show model choices in the quality menu. You see one row per kind:

RowWhat it covers
PlanningThe agent that plans and runs the work. It shows Wavemaker agent marked Default; there is nothing else to choose here.
ImagesStills, and edits that produce a still.
VideoGenerated clips, from text, images or references.
VoiceSpeech, dialogue and voice cloning.
Music & soundMusic, sound effects and sound for a video.

Select a row to open the model browser, titled Models · Images (or the kind you chose). In it you can:

  • search with Search models or providers;
  • switch kinds under Media type, each with a count of available models;
  • filter under How it’s paid: All, Wavemaker credits, Self-hosted or Your key;
  • filter by provider with All providers;
  • turn on Show unavailable to include models that can't run in your workspace right now, each with the reason.

Choosing a model applies it at once and leaves the browser open, so you can set the next kind. The footer shows what is selected. Use the default puts the current kind back on its default.

Auto

The first option in each media row is Auto. Auto leaves the choice to the agent and follows your quality level; the browser describes it as "Follows your quality setting, then the next model if one refuses or fails". With Draft, the Images row reads Auto · Gemini 3.1 Flash Image, because Draft starts images on that model.

Models you can pick

The browser lists every model Wavemaker can run for each kind, grouped by provider, including self-hosted and partner models, so the full list is long. The catalog's named models are below; Default marks the one new conversations start from.

KindNamed choices
VideoH3 · self-hosted (Default), LTX 2.5 · self-hosted, Gemini Omni 1.1 Flash, Kling 3.0, Runway Gen-4.5, Luma Ray 3.2, Seedance 2, Grok Imagine Video, MiniMax H3 · partner API, Vidu Q3
ImagesGemini 3 Pro Image (Default), Gemini 3.1 Flash Image, GPT Image, Grok Imagine Image, Seedream 5, Ideogram 3, Ideogram 4, Ideogram 4.5, Krea 2 · self-hosted, MOSS-Image · self-hosted
VoiceElevenLabs (Default), OpenAI speech, MiniMax speech
Music & soundElevenLabs music & sound, ACE-Step music · self-hosted

A model can be listed but unavailable at a given moment. The row then reads Unavailable, and the browser hides it unless you turn on Show unavailable.

What a pick means

  • A model you pick is a firm choice. For that kind of media in that message, the agent uses your model. If it can't, it doesn't switch silently: the call is refused and the agent stops to ask you.
  • A kind left on Auto is the agent's choice. For video, the agent is instructed to use Wavemaker's self-hosted H3 model by default, to move a shot to a partner video model only when you pick one, name one in your brief, or H3 failed for that shot, and to tell you which shots moved and why.
  • Saved defaults are starting points. Models from your default preset are where the agent starts. It may choose another when the brief needs it.
  • Adult (18+) stills always go to self-hosted Krea 2, whatever image model you picked. Adult work needs no plan and no Show 18+ material setting; the agent judges from the brief whether it asks for adult work. See Content policies.

If a model you chose is no longer offered, its row says it is no longer available and asks you to choose another model.

The agent

The Agent section of the quality menu names the hosted agent that plans and runs the work. When Wavemaker offers more than one, as it currently does with Claude, Codex and OpenCode, you choose among them there; Claude is the default. Your choice applies to a new chat. A conversation keeps its agent for its whole life: in an open conversation the menu shows its agent and says "To use a different agent, start a new chat."

How choices affect cost

Every model's use is charged in credits as it runs. The code makes these comparisons, and no others:

  • Draft usually costs less, and High quality usually costs more.
  • The standard presets describe their spend: Draft is Lower-cost exploration, Standard is Balanced spend · self-hosted video, and the two partner-video presets are Partner video · higher spend.

Prices for individual models aren't shown in the menu. To see what a message is likely to cost with your current choices, use Check what this message costs, below. See Credits for how charges settle.

Planning budget

Planning budget sits at the bottom of the quality menu. It sets the most one message may spend: the agent's model use and every tool call, generated media included.

  1. Open Planning budget. The chip shows Automatic when no limit is set.
  2. Type a whole number of credits under Total planning limit (credits).
  3. Send your message.

Sending reserves the limit. The reply stops when the limit is spent, and unused credits come back when it settles. Leave the field empty for Automatic: no limit for this message, though your balance and any account spending cap still apply.

If a reply reaches its limit, it stops and keeps what it made. A Needs you notice under the reply says what stopped it, for example "Stopped: the spending cap you set for this turn was reached after 40 of 40 credits.", then lists what is done and what is not. Select Continue to send a new message that picks up from the saved work, under the limit you set for it, or Stop here.

While a message is waiting to be accepted, its limit shows Locked with the pending request and can't be changed.

Cost check

Check what this message costs is in the same section. Write your message first, then select it.

The answer reads, for example: "About 12 credits for this message, up to 30 credits. Credits are charged as the work runs, not reserved up front. Checking does not spend credits." It can also show the assistant's share. If your balance is below the higher figure, it warns that the reply may stop before it finishes.

The check is an estimate based on your message and comparable past work. It doesn't set a limit and it isn't sent with your message. To cap spending, set a planning budget.

MessageMeaning
Write your message first, then check its cost.The brief is empty.
Sign in to check what this message costs.You aren't signed in.
The cost check returned an unexpected answer.Try again. Nothing was charged.

If something goes wrong

  • Checking available models… stays, or Model availability could not be checked. appears: select Retry. Auto still works and keeps your quality setting, so you can send without waiting. Use the default clears your picks.
  • Draft choices are not saved. Retry saving before sending.: your picks weren't stored with the draft. Change any setting again or reopen the menu, then send.
  • Not saved. Your choices still apply to the next message.: the preset didn't save, but the current message still uses what you chose.
  • You can save up to 30 presets.: update or delete one of your presets first.

Your first creative conversation · Write a useful creative brief · Credits

Related guidance