The following table lists the pricing of the current Partner Nodes. All prices are in credits.

​

Anthropic

​

Chat

Optional images add tokens billed at the same model input rates.

​

Beeble

​

Video: SwitchX Video Edit

​

Image: SwitchX Image Edit

​

BFL

​

Image

​

Image: FLUX 3 Image

Total credits = credits / image at the selected resolution.

​

Video: FLUX 3

Total credits = (credits / sec) × duration (duration is 5–20 seconds or auto).

​

Video: Flux Video Edit

Total credits = (credits / sec) × duration. Source clips of 0.7 to 15 seconds are edited from a text instruction.

​

Video: Flux Video Upscale

Rates are per output megapixel per second. Total credits = credits / MP / sec × output megapixels × duration. Output resolution ranges from 1080p (~2.1 MP) to 4K (~8.3 MP).

​

Bria

​

Image

​

Video

​

ByteDance

​

LLM: ByteDance Seed

Per million tokens. Rates below are for prompt length [0, 128]K; longer prompts use higher tiers. Optional images and videos add input tokens at the input rate.

​

Image: ByteDance Seedream

Per successfully generated image. Seedream 5.0 Pro uses resolution-based billing: 1K (≤2.61 MP, 9.50 credits) and 2K (>2.61 MP, 18.99 credits). Seedream 5.0 Flash bills a flat 5.43 credits per image at every resolution. The first image in each request is free. Each additional image beyond the first (including both output and input/reference images) adds 0.63 credits.

​

Image: ByteDance Image (deprecated)

​

Video: Seedance 2.5

Token-based billing. Estimated token consumption = (input video duration + output video duration) × output video width × output video height × output video frame rate ÷ 1024. When input is text, image, or audio only, input video duration = 0. Actual usage from the API is authoritative. Video input means a reference video is connected. Text, image, and audio references count as no video input. Minimum token consumption limits apply when the input includes video: if the estimated consumption is below the limit, billing uses the minimum. Reference videos must be at least 2 seconds long, and total input is limited to 30 seconds. Estimated charges for 16:9 output at 24 fps:

​

Video: Seedance 2.0

480p and 720p share the same rate. Fast and Mini do not support 1080p or 4K. Estimated token consumption: (Input video duration + Output video duration) × Output video width × Output video height × Output video frame rate ÷ 1024 When input is text, image, or audio only, input video duration = 0. Actual usage from the API is authoritative. Video input means reference video is connected. Text, image, and audio references count as no video input. For video-to-video inputs, a minimum token usage may apply. Seedance 2.0 outputs at 24fps.

​

Video: Seedance 1.x

Billed by tokens (same estimation formula as Seedance 2.0). Unlike Seedance 2.0, the token rate does not change by resolution; resolution changes how many tokens a clip uses (via width × height in the formula), so higher resolutions cost more. Token rate Approximate credits / 10 sec (scale with duration ÷ 10. Ranges reflect aspect ratio.) ByteDance Reference Images to Video (480p and 720p only):

​

Video: ByteDance vCube Video Enhance

Billed per second of enhanced output. The rates below are per second at 30 fps or below. Rates from 31 to 60 fps double the rate, and rates above 60 fps quadruple it. With source resolution or frame rate selected, the rate scales with the source video (approximately 1× to 4×).

​

Audio: ByteDance Seed Audio 1.0

Per minute of generated audio. Total credits = credits / min × duration (minutes).

​

Clarity AI

​

Image

Per megapixel of output. Total credits = (credits / MP) × output megapixels.

​

ElevenLabs

​

FishAudio

​

Magnific

​

Image: Upscale

Magnific Image Upscale (Creative) and Magnific Image Upscale (Precise V2) share the same billing. Upscaling runs in 2x steps (2x = 1 step, 4x = 2, 8x = 3, 16x = 4). After each step, pixel count ×4; each step is priced by the input megapixels at that step. Total credits = sum of all steps for your scale_factor. Example: 4x on a 1 MP image → step 1 at 1 MP (35.91) + step 2 at 4 MP (107.73) = 143.64 credits.

​

Image: Edit

​

Google

​

Chat: Google Gemini

Token-based billing; thought tokens use the output text rate.

​

Image: Nano Banana

Token-based billing.

​

Video: Google Gemini Omni

Billed per second of generated video. Rates are approximate.

​

Video: Google Veo

Total credits = (credits / sec) × duration.

​

HappyHorse

Total credits = (credits / sec) × duration. HappyHorse Video Edit supports HappyHorse 1.0 only. Billed at the same rates; total credits = (credits / sec) × output duration (3–15 sec, matches the input video).

​

HeyGen

Prices are set per-second or per-run. Avatar video bills per second at the rate of the selected avatar look’s type (Digital Twin, Photo Avatar, or Studio Avatar). Auto picks the engine per look, so its rate spans the full range above; a custom avatar ID is billed by its own type. HeyGen Video 1.0 clips run from 5 to 15 seconds and bill at the rate of the selected resolution. Reference to Video uses the no-reference rate until a reference image or video is connected, and each reference video adds 5 seconds to the billed duration. 2K output requires the 16:9 or 9:16 aspect ratio; auto works at 2K only when no reference media is connected.

​

Hitpaw

​

Video: HitPaw Video Enhance

Total credits = (credits / sec) × video duration. Rates depend on input FPS. Portrait Restore / General Restore (Portrait Restore 1x/2x, General Restore 1x/2x/4x): Ultra HD (Ultra HD Model 2x): Generative (Generative Model 1x; up to 4K):

​

Image: HitPaw General Image Enhance

Generative Portrait: Generative:

​

Ideogram

​

Krea

​

Kling

​

Video: Credits / sec

Total credits = (credits / sec) × duration, except Kling 3.0 Omni Edit Video and Kling Motion Control (billed per second of output; duration follows input video on Edit).

​

kling-video-o1

​

kling-v3-omni

​

kling-3.0-turbo

Always generates audio; the audio toggle is ignored.

​

kling-v3

​

kling-v2-6

​

kling-v2-5-turbo

Kling Text to Video, Kling Image (First Frame) to Video, and Kling Start-End Frame to Video use kling-v2-5-turbo in pro mode only, billed per run at a fixed 5s or 10s duration.

​

Kling Avatar 2.0

​

Video: Credits / run

​

Image

Total image credits = (credits / image) × n on Kling 3.0 Image; multiply by series_amount on Kling 3.0 Omni Image (kling-image-o1 always ×1).

​

Image: Virtual Try-On

Per run.

​

Lightricks

​

Video: Credits / sec

Total credits = (credits / sec) × duration. Resolution maps to 1280x720 / 720x1280 (720p), 1920x1080 / 1080x1920 (1080p), 2560x1440 / 1440x2560 (1440p), and 3840x2160 / 2160x3840 (4K). Durations over 10s on LTX-2.5 (Fast) require a 720p or 1080p resolution and 24/25 FPS; on LTX-2 (Fast) they require 1080p and 25 FPS.

​

Luma

​

Video

Flat /run prices match the node badge for each duration choice. Edit and reframe bill (credits / sec) × video length. ray-1-6 is always 105.50 / run. Luma Ray 3.2 Image to Video and Extend Video always bill the 5s price. Legacy Image to Video uses 5s or 9s from ray-2 / ray-flash-2 rows.

​

Image

Reference-image surcharge uses USD rounded to 4 decimals before converting to credits.

​

Meshy

Ultra Mode (available on Meshy 7.1, Meshy 7, and latest models) adds 60.3 credits: Text to Model 301.7 / run, Image to Model 422.4 / run (2K, 4K texture) or 482.7 / run (8K texture), Multi-Image to Model 422.4 / run (2K, 4K texture) or 482.7 / run (8K texture). The ultra pass resolution (2k, or 4k on Meshy 7.1 and latest) does not change the price. Meshy: Text to Model creates a draft mesh; run Meshy: Refine Draft Model on the returned meshy_task_id to texture it. Meshy: Image to Model and Meshy: Multi-Image to Model can texture in one step when should_texture is enabled.

​

Meta

​

Minimax

MiniMax Hailuo 02 Video only supports 6s at 1080P. MiniMax H3 bills per second (4–15s duration). MiniMax H3 Reference to Video adds 12.07 credits per reference image beyond 5, and bills reference videos at 27.16 credits/sec (768P) or 39.22 credits/sec (2K) (2s per video, up to 15s total, approximate). MiniMax H3 Context IR prices depend on attached media: text only, first/last frame, or reference media (approximate). MiniMax H3 Max bills per second (5–15s duration) at the rates above. MiniMax H3 Max Reference to Video billing has two parts: the output video is billed per second at the rates above, and reference inputs are billed as tokens. Tokens from every image, video, and audio clip in the request are pooled and share one free allowance:

Reference charge (credits) = max(0, total reference tokens - 4,096) × 6.03 / 1,000

The first 4,096 tokens per request are free; beyond that, each 1K tokens costs 6.03 credits. A reference video is tokenized at the output resolution, not at the source clip’s resolution: for a given output resolution, a 480P and a 768P clip of the same length cost the same. Images and audio are resolution-independent. Worked examples, each assuming no other reference input consumes the free allowance:

• A 1024 × 1024 image is 1,024 tokens: the first 4 images are free, then 6.03 credits each.
• A 2048 × 2048 image is 4,096 tokens: the first image is free, then 24.14 credits each.
• Generating a 5 s 768P video conditioned on two 1024 × 1024 images and one 5 s reference video: output 24.14 × 5 = 120.7 credits; reference tokens 2 × 1,024 + 37,296 = 39,344, minus the 4,096 allowance = 35,248 billable → 35.248 × 6.03 ≈ 212.6 credits. Total ≈ 333.3 credits.

​

OpenAI

DALL·E and Sora use fixed per-image or per-second rates. OpenAI GPT Image 2 and OpenAI ChatGPT bill from API token usage.

​

Image: DALL·E

DALL·E 2 supports 256×256, 512×512, or 1024×1024; total credits = (credits / image) × n. DALL·E 3 adds quality (standard / hd) and supports 1024×1024 or portrait 1024×1792 / 1792×1024.

​

Image: GPT Image

Billed from API token usage after generation. Approximate badge by quality (low / medium / high); actual cost follows token counts below. The deprecated OpenAI GPT Image 2 node (older implementation) uses the same models and rates. The GPT Image 2.5 models (gpt-image-2.5-flare, gpt-image-2.5-sunburst) share the same rates and add xhigh and max quality tiers. Output token counts scale with size and quality (for example, 1024x1024 at low is about 196 output tokens, roughly 1.86 credits). Each 1024x1024 reference image adds about 2.47 credits of image input, and prompt text is billed at the text-input rate. Multiplying images (n) scales the total.

​

Chat

Token-based billing. Badge rates are per 1K tokens; table is per 1M tokens. Optional images and files add input tokens at the same model rates.

​

Video: Sora

Total credits = (credits / sec) × duration (duration is 4, 8, or 12). sora-2 only supports 720x1280 and 1280x720.

​

OpenRouter

Badge rates are per 1K tokens; table is per 1M tokens. Final charges follow API usage.

​

Chat: OpenRouter LLM

​

Image: OpenRouter Image

Token-based billing. The badge estimates cost from the prompt length, reference count, and output size; the table lists the per-1M-token credit rates used by that estimate.

​

Image notes

• mai-image-2.6 and mai-image-2.6-flash bill from API token usage: the prompt, each reference image, and the generated image are tokenized, then charged at the rates above.
• Reference images are downscaled to at most 2048 x 2048 (about 4,096 tokens each) before upload.

​

Pixverse

Model v3.5 (fixed in the API; not selectable). 1080p is limited to 5s with normal motion. 8s clips only support normal motion (fast is coerced to normal).

​

PixVerse V6

PixVerse V6 nodes generate videos from 1 to 15 seconds with optional native audio. Total credits = (credits / sec) × duration. PixVerse V6 Fusion bills 2x the per-second rate for each second of reference video, instead of the output duration.

​

Quiver

​

Image

The retired arrow-1.1 models charged a fixed amount per run: arrow-2 and arrow-2-telos bill by the tokens a request consumes: Credits = (input tokens × input rate + output tokens × output rate) ÷ 1,000,000. Input tokens cover the prompt text and, for Image to SVG, the input image; output tokens cover the generated SVG. Both nodes bill the same rates. The node badge estimates a cost range from typical token counts; the final charge follows measured usage. reasoning_effort changes how many tokens the model spends before drawing: higher levels reason for longer, so they use more output tokens and cost more.

​

Qwen

​

Image

Per image. Output is billed per generated image; Qwen Image 3 Edit also bills 0.91 credits per input reference image (up to 3). Nodes with n bill (credits / image) × n. qwen-image-3.0-pro uses resolution-based output billing: 1K (≤2.25 MP) is 12.07 credits, 2K (>2.25 MP) is 22.63 credits. With the Edit node’s “auto” or “match input” size modes, Pro output pricing shows as a range (12.07–22.63); a custom size resolves to the fixed rate for that area.

​

Recraft

​

Image

Flat per-run pricing unless noted. Nodes with n bill (credits / image) × n. v3 generation nodes send model recraftv3 in the API request (not selectable in the node). The recraftv4_styles models always require a style_id or style reference images.

​

Reve

​

Image

Credits below are approximate base rates (test_time_scaling=1, upscale off). test_time_scaling > 1, upscale, and remove background can increase actual charges from the API. Fast edit/remix models use a flat base rate; upscale does not change the displayed estimate.

​

Rodin 3D

HighPack (addon_highpack) adds 168.8 credits on top of the Gen-2.5 base rate for that mode.

​

Runway

​

Video

Gen4 Turbo: 5s or 10s. Total credits = (credits / sec) × duration. Runway Aleph2 Keyframe and Runway Aleph2 Prompt Image are helper nodes with no separate charge; cost is on Runway Aleph2 Video to Video.

​

Image

​

Sonilo

Text to Music: total credits = (credits / sec) × duration (1–360 seconds).

​

sync.so

Total credits = (credits / sec) × duration.

​

Tencent

​

Image

HY Image 3.5 Preview bills one rate per generated image. 1K and 2K output, and custom sizes up to 4.19 megapixels, cost the base rate. 4K output is rendered at 2K and then upscaled by the model, and costs more, as does a custom size above 4.19 megapixels.

​

3D: Generation

3.1 uses the same rates as 3.0 but does not support LowPoly. Add-ons stack on the base rate: Multiview applies only to Hunyuan3D: Image(s) to Model when image_left, image_right, or image_back is connected.

​

3D: Post-processing

Hunyuan3D: 3D Texture Edit and Hunyuan3D: 3D Part each cost 126.6 / run.

​

Topaz

​

Image

Topaz Image Enhance (current node) charges per 8 megapixels of output (approximate): Topaz Image Enhance (Legacy) is deprecated. It has no fixed pre-run estimate; charge comes from the API estimate at poll time (API credits × 0.08 USD). Typical charges by input resolution:

​

Video

Approximate rates shown in the node are per source frame (not per second). Total credits ≈ (credits / src frame) × input frame count. With apo-8 interpolation enabled, add ~0.14 (1080p) or ~0.28 (4K) per source frame to the upscale rate. Upscaler Disabled with apo-8 interpolation only: the UI shows Interpolation only; charge comes from the API estimate at poll time. Topaz Video Enhance (Legacy) is deprecated. It has no fixed pre-run estimate; charge comes from the API estimate at poll time.

​

Tripo

Legacy generation prices are approximate. The node converts internal Tripo credits to USD at credits × 0.01.

​

Generation: v2.0 / v2.5

model_version v2.5-20250123. The v2.0-20240919 version has been retired and is no longer offered in the enum. texture or pbr counts as textured. On v2.5, geometry_quality: detailed does not change the price. Tripo: Text to Model Tripo: Image to Model and Tripo: Multiview to Model quad applies to Tripo: Text to Model and Tripo: Image to Model only. smart_low_poly adds +21.1 / run on every generation node in this section and caps face_limit at 20,000 (10,000 with quad).

​

Generation: v3.0 / v3.1

Same nodes with model_version v3.0-20250812 or v3.1-20260211. geometry_quality: detailed adds +42.2 / run on top of the texture rows below. Tripo: Text to Model Tripo: Image to Model and Tripo: Multiview to Model quad add-on (+10.55 / run) applies to Tripo: Text to Model and Tripo: Image to Model only. smart_low_poly adds +21.1 / run on every generation node in this section and caps face_limit at 20,000 (10,000 with quad).

​

Generation: P1

​

Post-processing

Tripo: Convert model charges 21.1 / run when quad is on, face_limit is set, texture_size ≠ 4096, texture_format ≠ JPEG, flatten_bottom or pivot_to_center_bottom is on, flatten_bottom_threshold ≠ 0, or scale_factor ≠ 1; otherwise 10.55 / run.

​

Multiview

The v3 API versions of Tripo: Image to Multiview and Tripo: Edit Multiview bill the same rates.

​

Generation: P2

The P2 nodes (Tripo P2: Text to Model, Tripo P2: Image to Model, Tripo P2: Multiview to Model) share one rate by texture. Prices are approximate.

​

Pruna

Pruna P-Video-2 bills per second of generated video. Total credits = (credits / sec) × duration (duration is 1-20 seconds or auto; when audio is connected, the video follows the audio length). Prices are approximate.

​

Vidu

​

Video: Q4

Total credits = (credits / sec) × duration.

​

Video: Q3

Total credits = (credits / sec) × duration. 2K is only on Vidu Q3 Image-to-Video Generation with viduq3-pro.

​

Video: Q2

Linear nodes (all except Multi-Frame): base + (credits / sec) × (duration − 1). Multi-Frame: frames × base + (credits / sec) × sum of segment durations (uses the Extension / Multi-Frame base column). Reference (viduq2): add +15.825 / run when audio is on. viduq2 linear bases above are for Text-to-Video. Reference-to-Video uses the same per-second rates but base 26.375 (720p) or 79.125 (1080p). viduq2-pro-fast, viduq2-pro, and viduq2-turbo linear bases are for Image-to-Video and Start/End Frame-to-Video. viduq2-turbo at 720p on those nodes: 8.44 at 1s (I2V), 10.55 at 2s (flat on I2V; Start/End minimum duration is 2s), then +10.55 / sec after 2s. The 5.275 / sec rate applies only to Extension and Multi-Frame, not I2V or Start/End.

​

Video: Q1

Fixed 5s at 1080p.

​

Wan

​

WaveSpeed

​

xAI

​

Image

​

Video

​

Cloud GPU