Gemini Omni Flash

Multi-modal Google model driven entirely from its own reference panel — images, a video clip, saved Characters and Voices, no frame cards.

Duration

4, 6, 8 or 10s

Output tiers

720p / 1080p / 4K

Aspect

16:9 or 9:16

VEO

3.1 Quality: Google's best-looking cinematic tier, priced flat per generation rather than per second.

Duration

4, 6 or 8s

Quality

Preview / 1080p / 4K

Inputs

Start + end frame

• 3.1 Quality
• 3.1 Fast
• 3.1 Lite

Seedance

2.5: The longest clips in the studio, audio generated with the video, and by far the widest reference set — but 480p/720p only, a lower ceiling than Seedance 2.0.

Duration

4–30s

Quality

480p / 720p

References

30 images · 10 videos · 10 audio

• 2.5
• 2.0
• 2.0 Fast
• 2.0 Mini

Kling

3.0: The most capable Kling — characters, synced sound, multi-shot, and a 4K tier.

Duration

3–15s

Quality

720p / 1080p / 4K

Inputs

Start frame; end frame outside multi-shot

• 3.0
• 3.0 Turbo
• 2.6
• 2.5 Turbo
• +1 more

Grok Imagine (video)

Strong motion, text and image-to-video, and the cheapest per second in the studio.

Duration

6–30s

Quality

480p / 720p

Inputs

Start frame

MiniMax H3

Two quality tiers — 768P and native 2K — first-to-last frame, and image, video and audio references.

Duration

5–15s

Quality

768P / 2K

Inputs

Start + end frame, or 9 images / 3 videos / 3 audio

HappyHorse 1.1

Text, image or multi-reference to video — images only, no video or audio references.

Duration

3–15s

Quality

720p / 1080p

Inputs

Start frame, or reference images

Wan

2.7: Four modes in one — text-to-video, image-to-video, video edit and reference-to-video — with driving audio, prompt expansion and negative prompts.

Duration

5, 10 or 15s

Quality

720p / 1080p

Inputs

Start frame, end frame, reference video, driving audio

• 2.7
• 2.6
• 2.5

MMAudio v2

Generates a foley track for a clip that has none.

Output

Audio muxed onto the source clip

Luma Ray-2 Modify

Restyles an existing clip end to end.

Output

The whole source clip, restyled

7Image

7ART's flagship image preset — our best image quality, pinned to the top of the same picker.

Resolutions

1K / 2K / 4K

References

Up to 16

Nano Banana

2: Up to 14 reference images — the safest pick for keeping a face consistent.

Resolutions

1K / 2K / 4K

References

Up to 14

• 2
• Pro

Seedream

5 Lite: Cheap and quick.

Resolutions

2K / 3K

References

Up to 10

• 5 Lite
• 4.5
• 4.0

GPT Image 2

Up to 16 references, strong at following instructions and at text inside the image.

Resolutions

1K / 2K / 4K

References

Up to 16

Z-Image

Cheap text-to-image plates — it takes no reference images at all.

Output

2K, single fixed quality

References

None — text-to-image only

Grok Imagine (image)

Text and image-to-image; one flat charge returns a whole batch.

Output

2K, single fixed quality

References

1

Batch

6 images in speed mode · 4 in quality

FLUX

FLUX.2 Pro: Top realism and typography.

Output

~1 MP

• FLUX.2 Pro
• FLUX.2
• FLUX.2 Turbo
• FLUX.2 Flash
• +8 more

Clarity Upscaler

Upscales a single image, 2× by default.

Input

Exactly 1 image

IC-Light v2

Relights a single image.

Input

Exactly 1 image

ElevenLabs

Turbo 2.5: The fast text-to-speech tier.

Speed

0.7×–1.2×

• Turbo 2.5
• Multilingual v2
• Eleven v3
• v3 Dialogue
• +4 more

HeyGen·2 models

HeyGen

Video Translate — Precision: Translates a video with a cloned voice and a re-synced mouth.

• Video Translate — Precision
• Video Translate — Speed

Sonilo v1.1

Sound design that lands on the action — foley or a music score synced to a video, or either one from a text description.

Modes

Video → sound effects · Video → music · Text → sound · Text → music

Sections

≥5s each, contiguous, ≤200 characters

Billing floor

15-second minimum

HeyGen

HeyGen Avatar IV

Photo to talking avatar.

Quality

720p / 1080p

Max length

60s

Volcengine Lip Sync

Re-lipsyncs an existing video to new audio.

Input

A source video

Max length

60s

Meshy 6

The best all-round mesh — quad topology, PBR materials and posing.

Modes

Image · Multi-view · Text

Polycount

10,000–300,000

Extras

PBR · quad topology · A-pose / T-pose

Hunyuan 3.1 Pro

The highest-detail mesh, with multi-view input.

Modes

Image · Multi-view · Text

Polycount

40,000–1,500,000

Extras

PBR

Trellis 2

Fast and affordable, from a single image.

Modes

Image only

Polycount

10,000–300,000

Images

1

Tripo 2.5

The fastest mesh in the 3D studio.

Modes

Image · Multi-view · Text

Polycount

10,000–300,000

Extras

PBR

Retexture Engine

Re-textures a mesh you already have, from a prompt or a style image.

Input

An existing mesh + 1 style image

Extras

PBR

Remesh

Remesh, rescale and re-export a model for printing or another pipeline.

Input

An existing mesh

Polycount

10,000–300,000

Formats

GLB · FBX · OBJ · STL · USDZ