Skip to main content
Radar

Every change to the models you run, with its source and its date. Releases, price changes, retirements, API changes, and incidents.Every change to the models you run, with its source.

Follow model changes
BenchLM
Image, video, music and audio APIs

AI media API pricing, per unit the provider bills

Per-image, per-second, per-song and per-minute prices for image, video, music, speech and realtime audio models on the OpenAI and Gemini APIs, each with its provider source and read date. Compared only inside one modality and unit.

Latest provider read

Product pages

Sora 2Video generation · 8 variantsAs of 2026-09-11, Sora 2 prices on the OpenAI API run from $0.05 per second (Sora 2, batch · 720p) to $0.70 per second (Sora 2 Pro, 1080p) across 8 published variants.Read 2026-09-11 Veo 3.1Video generation · 8 variantsAs of 2026-09-11, Veo 3.1 prices on the Gemini API run from $0.05 per second (Veo 3.1 Lite, lite · 720p) to $0.60 per second (Veo 3.1, 4k) across 8 published variants.Read 2026-09-11 Nano BananaImage generation · 8 variantsAs of 2026-09-11, Nano Banana prices on the Gemini API run from $0.0195 per image (Gemini 2.5 Flash Image (Nano Banana), batch) to $0.24 per image (Gemini 3 Pro Image (Nano Banana Pro), 4K) across 8 published variants.Read 2026-09-11 LyriaMusic generation · 3 variantsAs of 2026-09-11, Lyria prices on the Gemini API run from $0.04 per song (Lyria 3 Clip Preview, preview · 30-second clip) to $0.08 per song (Lyria 3.5, full song) across 3 published variants.Read 2026-09-11

Video generation

Video generation: every published row with its variant, price per unit and read date.
ProductVariantPriceRead
Sora 2720p$0.10 per second2026-09-11
Sora 2 Pro720p$0.30 per second2026-09-11
Sora 2 Pro1024p$0.50 per second2026-09-11
Sora 2 Pro1080p$0.70 per second2026-09-11
Sora 2batch · 720p$0.05 per second2026-09-11
Sora 2 Probatch · 720p$0.15 per second2026-09-11
Sora 2 Probatch · 1024p$0.25 per second2026-09-11
Sora 2 Probatch · 1080p$0.35 per second2026-09-11
Veo 3.1720p$0.40 per second2026-09-11
Veo 3.11080p$0.40 per second2026-09-11
Veo 3.14k$0.60 per second2026-09-11
Veo 3.1 Fastfast · 720p$0.10 per second2026-09-11
Veo 3.1 Fastfast · 1080p$0.12 per second2026-09-11
Veo 3.1 Fastfast · 4k$0.30 per second2026-09-11
Veo 3.1 Litelite · 720p$0.05 per second2026-09-11
Veo 3.1 Litelite · 1080p$0.08 per second2026-09-11

Text to speech

Text to speech: every published row with its variant, price per unit and read date.
ProductVariantPriceRead
TTS-1standard$15.00 per 1M characters2026-09-11
TTS-1 HDstandard$30.00 per 1M characters2026-09-11

Speech to text

Speech to text: every published row with its variant, price per unit and read date.
ProductVariantPriceRead
GPT Transcribestandard$0.0045 per minute2026-09-11
Whisperstandard$0.006 per minute2026-09-11

Realtime audio

Realtime audio: every published row with its variant, price per unit and read date.
ProductVariantPriceRead
GPT-Live-1standard$0.05 per minute2026-09-11

Image generation

Image generation: every published row with its variant, price per unit and read date.
ProductVariantPriceRead
Gemini 3.1 Flash Image (Nano Banana 2)0.5K$0.045 per image2026-09-11
Gemini 3.1 Flash Image (Nano Banana 2)1K$0.067 per image2026-09-11
Gemini 3.1 Flash Image (Nano Banana 2)2K$0.101 per image2026-09-11
Gemini 3.1 Flash Image (Nano Banana 2)4K$0.151 per image2026-09-11
Gemini 2.5 Flash Image (Nano Banana)standard$0.039 per image2026-09-11
Gemini 2.5 Flash Image (Nano Banana)batch$0.0195 per image2026-09-11
Gemini 3 Pro Image (Nano Banana Pro)1K/2K$0.134 per image2026-09-11
Gemini 3 Pro Image (Nano Banana Pro)4K$0.24 per image2026-09-11

Music generation

Music generation: every published row with its variant, price per unit and read date.
ProductVariantPriceRead
Lyria 3.5full song$0.08 per song2026-09-11
Lyria 3 Clip Previewpreview · 30-second clip$0.04 per song2026-09-11
Lyria 3 Pro Previewpreview · full song$0.08 per song2026-09-11

How these rows are kept honest

  • USD only, exactly as the provider publishes it; no currency conversion.
  • One row per priced unit variant (tier, resolution, duration mode). Never average variants into one price.
  • Compare rows only inside the same modality and unit. A per-second video price is never converted into a per-image or per-token figure.
  • priceUsd is null only with a note explaining why; a missing published price is unavailable, never zero.
  • Every row carries the provider page it came from and the date it was read.

Questions

How is AI image and video pricing different from token pricing?

Image, video and music models bill per generated output or per second, not per token. Sora 2 and Veo 3.1 are priced per second of video, the Nano Banana models per image, Lyria per song or clip, and speech models per character or minute. This hub keeps each modality in its own unit and never converts between them.

Which AI video API is cheapest per second?

From the rows on this hub, Veo 3.1 Lite at $0.05 per second of 720p and Sora 2 in batch at $0.05 per second of 720p are the lowest published rates, followed by Veo 3.1 Fast and Sora 2 standard at $0.10. The comparison is per second at the same resolution and says nothing about quality or length limits.

How often do media prices change?

Preview and promotional rates change without notice, and preview models can be retired. Every row here shows the date we read the provider page, rows are re-read before publication, and the registry contract records a missing published price as unavailable rather than zero.