Guides
Sume guides: the API, Formats, agents, models, media tools and pricing, each answer dated and sourced.
Sume TTS 409 tts_voice_language_mismatch: confirm and retry safely
A Sume TTS 409 means the voice's language differs from your request. Nothing was charged. Ask the user, then retry with confirm_language_mismatch true.
Ask Studio says my Short drags: cut the dead air in Python
When feedback says a Short drags, find pauses from word timings and rebuild it without them. Tested Python and the Timeline 1.0 body.
Announce a new Shorts episode with a YouTube post and a video frame
Make the still for a YouTube post that points to your next Shorts episode: pull the frame with Sume video frames and keep within the post limits YouTube lists.
Gemini 3.8 Flash TTS vs Flash-Lite TTS: what one hour of audio costs
One hour of speech is 90,000 audio tokens. Flash TTS is $0.81 and Flash-Lite TTS $0.54 at the 2026 standard rate, doubling in 2027. All tiers compared.
Claude Code 2.1.290 claude -p skips resources/list: Sume is tools-only
Claude Code 2.1.290 no longer holds the first claude -p turn for HTTP MCP servers to answer resources/list. Sume declares tools only, so nothing is lost.
Gemini 0.63.0 auth loop fix for headless keyring: Sume key from env
Gemini CLI 0.63.0 fixed an infinite auth loop from a headless keyring. In CI, keep Sume's key in an environment variable and one header, not a keyring.
Loading states for a Generate clip button: pending to cancelled
Map Sume's five video job statuses to five UI states, handle the 409 job_not_completed case, and poll at the interval the API suggests instead of a fixed timer.
Spin-off for an AI series: reuse a character sheet in a new show
A spin-off keeps one character and changes the world. Generate a new sheet with up to 16 references, then reuse the same stills across the new show's episodes.
Is my Short silent? Check has_audio before detach or captions
Video inspect with frames off returns probe.has_audio. Check it before audio detach or speech captions, which refuse a clip with no audio track.
Sume TTS Router: send sonic-3.6, sonic-latest or sonic-preview?
The Sume TTS Router has five Sonic ids. sonic-3.6 is stable, sonic-latest resolves to it, and sonic-preview is a beta that rejects pro voice clones.
Write an instrumental ad bed prompt for Lyria 3.5 on Sume Music
A prompt pattern for a no-vocals music bed on Sume Music: 1 to 5,000 characters, no duration field, $0.125 per generation. What to put in the prompt.
Gemini custom voices: 200 per project, 1-year expiry, 7-day voice keys
Gemini 3.8 TTS stores up to 200 prompted or replicated voices per project. Voices expire after a year and voice keys after 7 days. What to plan for.
Sume TTS word timestamps and sentence slices: the 70 ms boundary rule
Ask Sume TTS for timestamps.words and segmentation.mode sentence to get gapless sentence segments. The cut lands 70 ms after the last word unless you change it.
Animate a photo in a Short: why Timeline stills stay static
A still in a Timeline 1.0 slot is a static hold and its motion field is ignored. To move a photo, generate a 3-second Omni image-to-video clip and cut that in.
Sume TTS output: mp3, wav or raw, picked by the next step
Sume TTS defaults to mp3 at 44.1 kHz and 128 kbps; wav and raw cover editing and slicing. Pick the container, rate and encoding by what you do next.
Gemini Omni Flash 1.1 API: a vertical 4K clip with curl on Sume
Request a 9:16 4K clip from Gemini Omni Flash 1.1 on Sume: model id, resolution 4K, 3 to 10 seconds, audio always on, and the $3.00 price for 8 seconds.
Seedance 2.5 API in Python: a 30-second clip with an audio reference
Submit a 30-second Seedance 2.5 job on Sume with requests: an image and an audio reference in input_references, polling, download, and limits to check.
21:9 cinematic AI video API: which Sume models accept ultrawide
Seedance 2.5, Seedance 2.0, MiniMax H3 and H3 Max accept 21:9 on Sume. Kling, Wan, Omni and Grok do not. A 5-second H3 Max 768p clip is $0.50.
Freeze the last 1.5 seconds of a Short for an end card
Hold the final beat of a Short as a still: pull the frame at duration minus 1.5 s with Sume video frames, then lay it as a still slot in Timeline 1.0.
Sume price formula: list x 1.25 plus 5.5% fee, worked on six clips
How to predict a Sume bill: provider list price x 1.25 rounded up, then a 5.5% fee floored. Six video clips worked from list cost to hold, in micro-dollars.
Estimate a Gemini TTS bill from seconds of audio (Python, 25 tok/s)
A 20-line Python estimator for Gemini 3.8 Flash and Flash-Lite TTS audio output cost from seconds, at 25 tokens per second and Google's 2026 and 2027 rates.
Luma Ray3.2 or Kling 3 for a start-and-end-frame shot? Sume's option
Ray3.2 takes up to 16 keyframes and 20-second 1080p clips. On Sume, Kling 3 takes a first and a last frame for 4 to 15 seconds; Ray is not in the catalog.
Seedream 4.0 vs 4.5 vs 5.0 Lite on Sume: price, ratios, model ids
Three Seedream rows are listed on Sume: 4.0 at $0.0325, 4.5 at $0.05, 5.0 Lite at $0.04375 billed. Ratios, ids, edit support and the 4-image cap compared.
Duck music under a voiceover on Sume Timeline: duck_db, gain, fade
Sume Timeline 1.0 soundtracks take gain_db, loop, fade_out_seconds up to 10 and duck_db from 0 to 20. Settings for a voiceover over a music bed.
Reel series: keep one product's look across five AI video episodes
How to keep one product looking the same across a five-episode Reel series by pinning a frame or passing references on Sume, and which models accept which.
fastmcp-remote: connect a stdio-only MCP host to Sume's hosted MCP
FastMCP 3.4.0 added fastmcp-remote, a bridge from stdio-only hosts to HTTP servers with OAuth for HTTPS. Point it at mcp.sume.com/mcp, verify with mcp_health.
Runway API vs Sume: price of a Wan 3 and Seedance 2.5 10-second clip
Runway lists Wan 3 at 5 to 20 credits a second and Seedance 2.5 at 20 to 68, at $0.01 a credit. Here is the 10-second clip price next to Sume's, before the fee.
MCP Skills extension (SEP-2640): what Sume's hosted MCP exposes
SEP-2640 serves Agent Skills as skill:// MCP resources. Sume's hosted MCP declares tools only, so skills/list is not a Sume surface; here is what to use.
Ten 2,000-character scripts: Eleven v4 sale vs list vs Sume TTS
Eleven v4 is $0.022 per 1K characters on sale until Oct 12 and $0.08 at list. Ten 2,000-character scripts cost $0.44, $1.60, or $1.00 on Sume TTS.
Sora 2 vs Grok Imagine video 1.5: text-first or image-first
Sora took a text prompt. grok-imagine-video-1.5 on Sume requires an image_url, runs 4 to 15 seconds at 480p or 720p, and bills $0.0125 per second.
image_size or aspect_ratio on Sume: which wins and who accepts it
On Sume's Images API image_size takes priority over aspect_ratio. Custom pixels work on GPT, Seedream, Flux, Qwen and Recraft; size rejects WxH.
Node test contract test for a Sume video wrapper, no mock library
Test a Sora-replacement wrapper with node:test and a local http server that answers 202, in_progress, then completed. No network and no key needed.
Sign and verify a Sume webhook offline: a Python test vector
Build a sume-v1 test signature with Python's hmac and check your verifier against it, including an empty secret, an old timestamp and a rotation header.
TikTok 'Reference to Video' vs Sume frame_images for product moments
TikTok's Symphony Creative Studio lets advertisers prompt exact images and products at set moments. How the job maps to Sume frame_images and references.
fal vs OpenRouter vs Sume: three webhook shapes for one Sora port
Porting a Sora worker? fal takes webhook_url on a queued request, OpenRouter signs t=,v1= with a 5 minute window, Sume signs sume-v1 over timestamp.body.
Zed 1.23.2 optional MCP fields fix: read Sume's tools_schema anyway
Zed 1.23.2 fixed OpenAI model calls to MCP servers that treated optional fields as required. For Sume, tools_schema shows which fields are required.
TikTok AI voiceover: the disclosure log to keep for each video
TikTok now labels AI content and joined the C2PA steering committee. Keep a short per-video log for a synthetic voiceover: voice source, consent, label, date.
Cursor iOS remote control for local agents: keep paid Sume calls safe
Cursor can now show and answer local agents from its iOS app. If one holds Sume MCP, grant read-only, send dry_run first, and cap max_spend_usd.
Creator or Brand Content Series on TikTok: which AI pilot to render
TikTok's The Next Episode splits into Creator Content Series and Brand Content Series. Two different pilots, with Sume prices: $3.75, $7.74 and a $0.95 avatar.
Azure DevOps remote MCP is GA: run it beside Sume's hosted MCP
Azure DevOps remote MCP went GA in sprint 278 with Foundry and Copilot Studio support. Here is how to hold it and Sume's MCP with separate credentials.
Which Nano Banana models can you call on Sume? Google lists five
Google's Gemini API lists five Nano Banana models; Sume's image catalog lists two, Nano Banana 2.1 and Pro. Nano Banana 2 ids still run, as 2.1.
Map a Sume API HTTP status to the next step: one Python function
One Python function turns a Sume status, error code and headers into the next action: wait, retry with the same key, fix the request, or stop. Tested.
How do I caption a silent micro-drama for Shorts with overlay cues?
Send authored text as cues with start and end times to the Sume caption job. A silent clip fails speech captions with caption_no_speech, so use cues.
Is there a seed to keep Shorts characters the same across episodes?
No. No video model in Sume's catalog accepts a seed, and the docs say a seed is not deterministic on all providers. Use references and first frames instead.
Can I render a Shorts season on Sume's Free plan?
Yes, in waves: the Free plan runs 1 job at a time and accepts 6, then returns 429 queue_full. Ten 30-second episodes need two submit waves.
Keep a Shorts series consistent with Omni Flash video references
Gemini Omni Flash 1.1 on Sume accepts image and video references for 3 to 10 second clips. Feed the last episode's clip in as the next one's reference.
Does Wan 3.0 allow 2-second clips for a Shorts cold open?
Yes. Wan 3.0 on Sume accepts 2 to 30 seconds per call, so a 2-second cold open at 720p is $0.25, and it is the cheapest hook test in a season.
Sume queue_full 429 on a Shorts batch: how do I retry it?
queue_full means your accepted-job capacity is full: 6 on Free, 24 on Pro. Wait for jobs to finish, then resubmit with the same Idempotency-Key.