Dual Mode

Dual Mode lets you run two AI models simultaneously from a single Generate click. Instead of generating with one model, evaluating, then trying another, you get both results at the same time — ready for direct comparison.

How It Works

Select a primary model from the model dropdown as usual
Enable Dual Mode using the toggle in the Generate tab
Select a secondary model from the second dropdown that appears
Click Generate — both models run in parallel using the same prompt and source media

Both results download and appear as selectable cards in the preview area. Click one to preview it, then import it — the other card auto-selects so you can decide on the second result immediately.

Compatibility Check

Before generation starts, modelBridge compares the input schemas of both models to identify shared fields. Fields that represent the same concept but use different names — like prompt vs multi_prompt, or aspect_ratio vs video_size — are matched through an alias resolution system.

The compatibility summary shows:

Number of shared inputs — fields that both models understand
Unmatched fields — fields that exist in one model but have no equivalent in the other

If the secondary model has a required field with no match in the primary model’s form, generation is blocked with a specific warning explaining which field is missing.

Cost Preview

The cost badge shows the combined estimated cost for both models. Primary model parameters are read directly from the form; secondary model parameters are normalized to valid values within that model’s constraints before calculation.

The display adapts to the pricing data available:

Both models have USD pricing → ~$X.XX · 2 models with per-model breakdown
Both use token-based pricing → token · 2 models
Mixed pricing types → $X.XX + token · 2 models
One or both unavailable → shown honestly

A cost confirmation dialog appears when the combined cost exceeds $8.00.

Same-Source-Clip Handling

Both Dual Mode models use the same source clip from your timeline. This matters when importing results:

When you import the first result using Replace on Timeline, the source clip is replaced and marked internally
The second result automatically falls back to Insert at Playhead since the original source clip no longer exists on the timeline
This prevents conflicts — you never have to think about which result replaced what

Selecting and Importing Results

Results appear as clickable cards with a highlighted border on the active selection. The workflow is:

Both results arrive → first card is auto-selected
Review the selected result in the preview area
Click Import to Timeline or Save to Project Bin
The remaining card auto-selects for your next decision
Import or discard the second result

Each result gets its own contextual import text based on its model type, so you know exactly what the import action will do.

When to Use Dual Mode

Comparing model quality — test Kling v3 against Wan 2.6 on the same prompt to see which produces better results for your specific use case
Comparing speed vs quality — run a fast model (LTX) alongside a slow model (Kling Pro) to see if the quality difference is worth the wait
Comparing voices — for TTS models, generate the same script with two different voice models side by side
Client presentations — generate two options to present without doubling your generation workflow

Limitations

Both models always use the same source clip — you cannot assign different sources to each slot
The preview does not indicate which source clip was used (it is always the same one)
Cost history logs dual generations as two independent entries with no flag linking them as a pair
Only Save to Project Bin is available as an action — timeline import for both slots in a single action is not supported yet
Parameters that exist only in the secondary model use that model’s defaults — they are not configurable from the primary model’s form