Pika API Club: 100+ models at reduced pricing, questions to ask first
Pika's API Club (Aug 5) offers 100+ models at reduced pricing, and the Sep 17 relaunch adds audio models. Seven questions to ask an aggregator.

Pika's blog, read 2026-10-03, describes API Club (August 5) as access to 100+ models at reduced pricing, and the post "Welcome to the New Pika" (September 17, 2026). It also lists audio models named Soundtrack, Music, SFX and Speech from August. Before you move a pipeline onto any aggregator, get answers to seven questions.
What Pika published
Only these items come from the Pika blog.
| Date | Item |
|---|---|
| Aug 5, 2026 | API Club: 100+ models at reduced pricing |
| Aug 2026 | Audio models: Soundtrack, Music, SFX, Speech |
| Sep 17, 2026 | Welcome to the New Pika |
Seven questions for any aggregator
These apply to Pika, to Sume and to everyone in between.
- Reduced compared with what: the model owner's list price, or the aggregator's own earlier price?
- Is there a margin on top of provider list? On Sume the margin is stated: provider list times 1.25 on every video model.
- Can you list the catalog and its limits by API? Sume returns
supported_durations,supported_resolutionsandpricing_skusfromGET /v1/videos/models. - What happens when a provider is down or at capacity? Sume documents
provider_capacity_exceededand tells you to retry with the same idempotency key. - Are retries safe? Sume replays an
Idempotency-Keyas the original job. - Which model answered? With
sume/auto, Sume does not disclose the family that served the request; if you need to know, pin a catalog id. - How do you export your jobs and outputs if you leave?
The audio side
Pika's audio line-up, music and sound effects and speech, mirrors a normal video pipeline. On Sume, music goes through the Music Router (POST /v1/music-router/generate, sume/music-auto by default, Lyria 3.5 today), speech through the hosted MCP tts_create tool, and sound effects are not covered by the docs reviewed here, so do not plan on them without checking the catalog.
Sume's Music Router guide says every router model charges the fixed Music price per audio generation, and duration fields are rejected, so you steer length in the prompt.
Reading a reduced price
A reduced price is a claim about a baseline, so find the baseline. If the comparison is against the owner's list, ask whether the owner's list includes volume terms. If it is against the aggregator's earlier price, it says little about what you will pay compared with going direct.
Sume states its approach plainly: workspace USD balance, reserved at submit, provider list times 1.25. That gives you a number you can verify against the provider's own page.
Decide on a pilot
Run the same ten prompts through each candidate, record cost per finished clip, failure rate and time to result, then compare. A reduced price matters only if the output and the retry behavior meet your standard.
Sources
Related posts
More in Comparisons
- Pocket TTS self-hosted vs hosted TTS: a break-even formula
Kyutai Pocket TTS is MIT-licensed but needs a server. Cost per audio hour is server price divided by speed. On Sume it is $0.0475 per 1,000 characters.
- 30-minute podcast: transcript plus summary voiceover cost by vendor
Transcribing 30 minutes and reading a 4,500-character summary aloud costs $0.12 on xAI, $0.20 on OpenAI, $0.21 on ElevenLabs and $0.51 on Sume. Math inside.
- One podcast episode's audio bill: MAI, Lyria and transcript vs Sume
A 5,000-character intro, one music track and a 60-minute transcript cost $0.695 on MAI Flash, Lyria 3.5 and MAI-Transcribe, and $0.9625 on Sume.
- Powtoon Lite's 10-minute cap vs a Sume 30-minute timeline render
Powtoon caps video at 10 minutes on Lite, 20 on Professional and 30 on Advanced. A Sume Timeline render accepts up to 1,800 seconds, on any plan.
Written by Sume